Trans-ancestral genome-wide association studies of brain imaging phenotypes

Trans-ancestral genome-wide association studies of brain imaging phenotypes | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Trans-ancestral genome-wide association studies of brain imaging phenotypes Chunshui Yu, Jilian Fu, Quan Zhang, Jianhua Wang, Meiyun Wang, and 39 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-2047527/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 29 May, 2024 Read the published version in Nature Genetics → Version 1 posted You are reading this latest preprint version Abstract Genome-wide association studies of brain imaging phenotypes are mainly performed in European populations, but other populations are severely under-represented. Here, we conducted Chinese-alone and trans-ancestral genome-wide association studies of 3,414 brain imaging phenotypes in 7,058 Chinese and 33,224 European individuals. We identified 37 novel variant-phenotype associations in Chinese-alone analyses and 459 additional novel associations in trans-ancestral meta-analyses under the thresholds of P < 1.46 × 10 − 11 for discovery and P < 0.05 for replication. We pooled genome-wide significant associations for brain imaging phenotypes identified in either single-ancestral or trans-ancestral analyses into 6,361 independent significant associations. These associations were unevenly distributed in the genome and across the brain phenotypic subgroups and demonstrated significant enrichment for nervous system development and signal transduction. We further categorized the 4,890 pooled genome-wide significant associations whose index variants were included in both Chinese and European analyses into 43 ancestry-specific and 3,524 ancestry-shared associations. Loci of the 6,361 pooled genome-wide significant associations for brain imaging phenotypes were shared by 16 brain-related non-imaging traits including cognition, personality, risk behavior, addiction, and neuropsychiatric disorders. Our results provide a valuable catalog of genetic associations for brain imaging phenotypes in diverse populations. Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Introduction Brain imaging phenotypes reflect the structure, function and connectivity of the brain, and they are indicative of cognitive performance and vulnerability of neuropsychiatric disorders 1 . Quantitative brain imaging phenotypes show high heritability 2 and thus the investigation of genetic architecture of the human brain will shed light on the causes for individual differences in cognitive processing and mechanisms of neuropsychiatric disorders. In the past decades, many neuroimaging genetics studies have explored the genetic associations of the brain structure and function, of which several large-scale genome-wide association studies (GWASs) in individuals of European ancestry (EUR) provide unbiased insight into the genetic architectures of brain imaging phenotypes 3 – 7 . The human brain structure, function and connectivity and the prevalence of inherited brain disorders may differ by ancestries. For example, there are significant differences in regional cortical volume, thickness, and surface area between individuals of East Asian ancestry (EAS) and EUR 8 – 10 . In central nervous system demyelinating disorders, Caucasians show higher incidence of multiple sclerosis 11 but lower incidence of neuromyelitis optica spectrum disorder 12 than Asians. Although non-genetic factors may contribute to the differences in brain properties and disorders among ancestries 13 , genetic architectures such as effect size, linkage disequilibrium (LD), and allele frequency (AF) contribute a lot to these differences among ancestries 14 . Previous GWASs have successfully identified thousands of genetic associations with brain imaging phenotypes 3 – 7 , 15 , yet non-EUR populations are severely under-represented, which prevents us to differentiate ancestry-specific from ancestry-shared genetic associations with brain imaging phenotypes. The knowledge may improve our understanding of ancestry-shared and ancestry-specific relationships between genetic variations and brain disorders since the brain structure and function are intermediate phenotypes linking genetic variations to neuropsychiatric disorders. Despite trans-ancestral GWASs could discover novel variant-phenotype associations and distinguish ancestry-specific from ancestry-shared associations 16 , 17 , trans-ancestral GWASs for brain imaging phenotypes are still lacking. The main obstacle to trans-ancestral GWASs of brain imaging phenotypes is the lack of large-scale non-EUR cohorts with both genomic and neuroimaging data, but this dilemma will be broken by the emergence of the Chinese Imaging Genetics (CHIMGEN) cohort 18 ( http://chimgen.tmu.edu.cn ). This cohort has collected genomic and neuroimaging data from 7,306 healthy Chinese Han participants, from which we generated 6,830,145 genomic variants and 3,414 brain imaging phenotypes that have been included in GWASs of EUR individuals from the UK Biobank (UKBB) dataset 15 . We performed EAS-GWASs for these brain imaging phenotypes based on CHIMGEN data and trans-ancestral GWASs based on GWAS summary statistics of the 3,414 brain imaging phenotypes from EAS (CHIMGEN) and EUR (UKBB) populations. Here, we were interested in: (a) detecting novel genetic associations by incorporating EAS population; (b) exploring the distribution and function of all loci associated with brain imaging phenotypes; (c) identifying ancestry-shared and ancestry-specific associations; and (d) investigating genetic sharing between brain imaging phenotypes and other brain-related traits such as cognition, personality, and neuropsychiatric disorders. A schematic summary is shown in Extended Data Fig. 1 . Results Genetic discovery in EAS-GWASs In 7,058 EAS (Chinese Han) participants from the CHIMGEN study, we conducted the first non-EUR GWASs for 3,414 brain imaging phenotypes at the 6,830,145 autosomal variants with a minor allele frequency (MAF) > 1%, in which genetic effects were estimated with respect to the number of copies of the non-reference allele. In the discovery dataset of 5,025 EAS participants, we identified 647 genome-wide significant associations ( P < 5 × 10 − 8 ) between genetic variants and brain imaging phenotypes (Fig. 1 a, Supplementary Table 1), and 295 associations were confirmed at P < 0.05 with the same direction of effect in the replication dataset of 2,033 EAS participants. Using a threshold of P < 1.46 × 10 − 11 to additionally correct for the 3,414 GWASs, we found 133 significant variant-phenotype associations in the discovery dataset (Fig. 1 a, Supplementary Table 2). Of the 133 associations, 124 were replicated at P < 0.05 with the same direction of effect in the replication dataset, 124 were also confirmed with a false discovery rate (FDR) of less than 0.05, and 63 survived the Bonferroni-adjusted significance threshold of P < 3.75 × 10 − 4 . These significant associations were unevenly distributed across the genome, of which the chromosomes 5, 11 and 12 showed more associations with brain imaging phenotypes (Fig. 1 a). Consistent with the reported pleiotropic variant of rs67827860 at 5q14.3 in EUR participants 4 , this variant also showed extensive associations with brain diffusion imaging phenotypes in EAS participants (Extended Data Fig. 2 ). In addition, we found another pleiotropic variant of rs111737551 at 11p11.2 that also showed significant associations with various brain diffusion imaging phenotypes (Fig. 1 b-d). This indel variant is located at an intron of CD82 , which regulates oligodendrocyte progenitor migration and white matter myelination 19 , and mediates age-related cognitive decline 20 . To test the potential of EAS-GWASs in identifying new associations that have not been reported in EUR-GWASs for the same brain imaging phenotypes at the same statistical thresholds, we defined a reference list of the known associations as those with P < 1.46 × 10 − 11 in the discovery dataset (22,138 EUR participants) and P < 0.05 in the replication dataset (11,086 EUR participants) in the currently largest EUR-GWASs for the 3,414 brain imaging phenotypes 15 , resulting in 1,478 associations (Supplementary Table 3). Here, we defined a novel association for EAS-GWASs if the locus of the index variant did not overlap with any loci of the same brain imaging phenotype in the reference list of the known associations. Of the 124 associations (discovered at P < 1.46 × 10 − 11 and replicated at P < 0.05) in EAS-GWASs, 37 (29.8%) associations were considered as novel by this definition (Fig. 1 a, Supplementary Table 4), indicating that GWASs conducted in the non-EUR populations (such as Chinese Han) can reveal novel variant-phenotype associations that are absent in EUR-GWASs even with a larger sample size. For example, rs77768175 at 12q24.13 demonstrated significant association with the right caudate volume (Fig. 1 e) only in EAS-GWAS. This polymorphic variant in EAS [MAF = 0.1607 in 1000 Genomes Project phase 3 (1KGP)] shows far less polymorphic in EUR (MAF = 0 in 1KGP), which may explain the absence of signal in EUR-GWAS. Besides, this is an intron variant of HECTD4 which has been associated with epilepsy in EAS but not in EUR 21 . Another EAS-specific association was the correlation of rs2274224 at 10q23.33 with functional activity amplitude of the ventral attention resting-state network (RSN) obtained from the independent component analysis (ICA) (Fig. 1 f). This missense variant affects PLCE1 protein coding and has been associated with migraine 22 . Genetic discovery in trans-ancestral GWAS meta-analyses At 5,950,889 autosomal variants included in both CHIMGEN and UKBB datasets, trans-ancestral GWAS meta-analyses were conducted based on the fixed-effect model for the 3,414 brain imaging phenotypes shared by the two datasets. In the discovery stage, trans-ancestral GWASs were performed based on summary statistics from the discovery stage of the EAS-GWASs (5,025 CHIMGEN participants) and EUR-GWASs (22,138 UKBB participants). In the replication stage, trans-ancestral analyses were conducted based on summary statistics from the replication stage of the EAS-GWASs (2,033 CHIMGEN participants) and EUR-GWASs (11,086 UKBB participants). In the discovery dataset (5,025 EAS and 22,138 EUR), trans-ancestral GWAS meta-analyses revealed 6,920 genome-wide significant variant-phenotype associations ( P < 5 × 10 − 8 ) (Supplementary Table 5), of which 5,065 associations were confirmed at P < 0.05 in the replication dataset (2,033 EAS and 11,086 EUR). Using a Bonferroni-adjusted threshold of P < 1.46 × 10 − 11 , we still found 1,746 significant variant-phenotype associations in the discovery dataset (Supplementary Table 6). Of these 1,746 associations, we confirmed 1,677 associations at P < 0.05 in the replication dataset, 1,675 at an FDR-adjusted threshold of q < 0.05, and 1,076 at a Bonferroni-adjusted threshold of P < 2.86 × 10 − 5 . Among the 1,677 variant-phenotype associations identified in the two-stage trans-ancestral GWAS meta-analyses (discovered at P < 1.46 × 10 − 11 and replicated at P < 0.05), 459 (27.4%) were considered as novel because the locus of the index variant did not overlap with any loci for the same brain imaging phenotype derived from either EAS-GWASs or EUR-GWASs (Supplementary Table 7). These novel associations were unevenly distributed across the genome and 32 subgroups of brain imaging phenotypes, of which chromosomes 3, 5 and 11 (Fig. 2 a and 2 b) and subgroups of cortical volume, surface area and white matter diffusion (Fig. 2 c and 2 d) showed more novel associations. An example is the novel association at 7q22.1 with the left cerebellum VIIIb volume (Fig. 2 e). The index variant rs1627052 is located at an intron of RELN , which regulates neuronal migration and cortical layering in the brain, and mutations of this gene are associated with autosomal recessive lissencephaly with cerebellar hypoplasia 23 . Another example is the novel association at 7p22.1 with the brain stem volume (Fig. 2 f). The index variant rs2640 is a missense mutation of EIF2AK1 , showing relations with developmental delay, leukoencephalopathy, and neurologic decompensation 24 . Pooled genome-wide significant associations We pooled all genome-wide significant associations (discovered at P < 5 × 10 − 8 and replicated at P < 0.05) for the 3,414 brain imaging phenotypes identified by any of the EAS-GWASs, EUR-GWASs, or trans-ancestral GWASs. We totally identified 6,361 independent associations after merging the overlapping genetic loci derived from the same brain imaging phenotype (Supplementary Table 8). Of these 6,361 associations, 1,952 were still significant at P < 1.46 × 10 − 11 with an additional correction for the 3,414 GWASs in the discovery dataset. These 6,361 associations were unevenly distributed across the genome with chromosomes 5 and 17 showing more associations (Fig. 3 a and 3 b). These associations spanned all subgroups of brain imaging phenotypes with the subgroups of cortical volume, surface area and white matter diffusion showing more associations (Fig. 3 c), even considering the different numbers of phenotypes in different subgroups (Fig. 3 d). For each subgroup of brain imaging phenotypes derived from the same brain atlas or parcellation, significant associations also showed uneven spatial distribution across the brain (Fig. 3 e-g and Extended Data Fig. 3 ). For example, we observed more associations in the calcarine sulcus, cuneus, precuneus and lingual gyrus for structural imaging phenotypes of the cerebral cortex (Fig. 3 e), in the forceps minor, superior and inferior longitudinal fasciculi and anterior thalamic radiation for diffusion imaging phenotypes (Fig. 3 f), in the salience RSN for functional activity amplitude (Fig. 3 g), and in the basal ganglion RSN for functional connectivity (Fig. 3 g). Genetic variants of the 6,361 independent significant associations were assigned to 856 protein-coding genes (Supplementary Table 9) based on the criteria of the location of a variant within 10 kb around a gene. Using all protein-coding genes (n = 20,589) as the background, we identified 39 functional enrichment terms in Reactome Pathway Database with an FDR-corrected q < 0.05 (Fig. 3 h and Supplementary Table 10). The term with the most significant enrichment was nervous system development (FDR-corrected q = 4.74 × 10 − 6 ), and the relevant pathways included axon guidance (FDR-corrected q = 4.81 × 10 − 6 ), the regulation of commissural axon pathfinding by SLIT and ROBO (FDR-corrected q = 5.29 × 10 − 3 ) and others. Another category of terms with significant enrichment was signal transduction (FDR-corrected q = 7.68 × 10 − 6 ) involved in neuronal differentiation and migration, such as CRMPs in Sema3A signaling for axonal outgrowth (FDR-corrected q = 2.34 × 10 − 2 ), Netrin-1 signaling for axon guidance (FDR-corrected q = 5.28 × 10 − 5 ), and NCAM signaling for neurite outgrowth (FDR-corrected q = 1.50 × 10 − 2 ). These pooled genome-wide significant loci provide a hitherto largest catalog of genetic associations for human brain imaging phenotypes in diverse populations. Ancestry-specific and ancestry-shared genetic discovery From the GWAS results of the same brain imaging phenotypes conducted in EAS and EUR 15 , both ancestry-shared and ancestry-specific associations were observed. Taking minimal diffusivity (L3) of the left anterior corona radiata as an example, this brain imaging phenotype was associated with 5q14.3 in both EUR and EAS (ancestry-shared), but with 11p11.2 only in EAS (EAS-specific) and with 17q21.31 only in EUR (EUR-specific) (Fig. 4 a). To systematically investigate the ancestry-shared and ancestry-specific genetic associations with brain imaging phenotypes, Cochran’s Q-test (CQ-test) was used to quantify the heterogeneity of effect size of the pooled 4,890 independent genome-wide significant associations (discovered at P < 5 × 10 − 8 and replicated at P < 0.05) whose index variants were included in both EAS-GWASs and EUR-GWASs. The CQ-test for discovery was conducted based on GWAS summary statistics from the discovery samples of EAS (n = 5,025) and EUR (n = 22,138) and the CQ-test for replication was performed based on GWAS summary statistics from the replication samples of EAS (n = 2,033) and EUR (n = 11,086). Ancestry-shared associations were defined as those of P ≥ 0.05 in CQ-tests for both discovery and replication and ancestry-specific associations were defined as those discovered at a Bonferroni-adjusted threshold of P < 1.02 × 10 − 5 and replicated at P < 0.05. From 4,890 variant-phenotype associations, we identified 3,797 homogeneous ( P ≥ 0.05) and 72 heterogeneous ( P < 1.02 × 10 − 5 ) associations between EAS and EUR in the discovery samples. Of these 3,797 homogeneous associations, 3,524 were replicated at P ≥ 0.05 and considered as ancestry-shared associations (Fig. 4 b and Supplementary Table 11). Of these 72 heterogeneous associations, 43 were replicated at P < 0.05 and considered as ancestry-specific associations (Fig. 4 b and Supplementary Table 11); 38 associations were still significant at an FDR-adjusted threshold of q < 0.05, and 10 at a Bonferroni-adjusted threshold of P < 6.94 × 10 − 4 . The brain imaging phenotypes from all three imaging modalities consistently showed more ancestry-shared associations than ancestry-specific associations, and the portion (1.76%) of ancestry-specific associations for functional imaging phenotypes was five times greater than that (0.31%) for diffusion imaging phenotypes (Fig. 4 b). To account for the bias due to unbalanced sample sizes between EAS and EUR, we also validated the identified ancestry-shared and ancestry-specific associations in EAS (n = 7,058 from CHIMGEN) and EUR (n = 8,428 from UKBB) 4 with comparable sample sizes. After excluding variant-phenotype associations absent in the EUR-GWASs 4 , we conducted CQ-tests for the remaining 1,845 ancestry-shared associations and 27 ancestry-specific associations. We observed 1,750 (94.85%) ancestry-shared associations at P ≥ 0.05 and 27 ancestry-specific associations (100%) at P < 0.05 (Fig. 4 c). Of the 27 ancestry-specific associations, all were also replicated at an FDR-corrected threshold of q < 0.05 and 25 at a Bonferroni-corrected threshold of P < 1.85 × 10 − 3 . In the structural imaging phenotypes, ancestry-specific associations were mainly observed in the bilateral putamen and precentral gyri, left pallidum and cerebellum VIIIb, and right fusiform gyrus and superior parietal cortex (Fig. 4 d). In the diffusion imaging phenotypes, ancestry-specific associations were presented in fiber tracts such as the bilateral superior cerebellar peduncles, genu of corpus callosum, right superior corona radiata, and left medial lemniscus (Fig. 4 e). In the functional imaging phenotypes, ancestry-specific associations were found in four functional connectivity involving seven brain RSNs derived from the ICA (Fig. 4 f). Three typical examples of ancestry-shared and ancestry-specific associations are presented in Fig. 4 g-i. Briefly, an EAS-specific association was uncovered between rs8081528 (17q12) and fractional anisotropy (FA) of the left medial lemniscus (Fig. 4 g). This variant is located at an intron of CDK12 , which regulates axonal elongation, neurogenesis, and neuronal migration 25 . An EUR-specific association was observed between rs2923402 (8p11.21) and left pallidum volume (Fig. 4 h). The rs2923402 is an intergenic variant close to CHRNB3 , which encodes subunits of nicotinic acetylcholine receptors and is associated with nicotine dependence 26 . An ancestry-shared association was identified between rs13164785 (5q14.3) and mean diffusivity (MD) of the left uncinate fasciculus (Fig. 4 i). This variant is located at an intron of VCAN , which encodes a large chondroitin sulfate proteoglycan, a major component of the extracellular matrix. The protein VCAN is involved in cell adhesion, proliferation, migration, and angiogenesis, and is critical for tissue morphogenesis and maintenance 27 . Colocalizations with brain-related non-imaging traits For the 6,361 pooled independent genome-wide significant associations (discovered at P < 5 × 10 − 8 and replicated at P < 0.05) of the 3,414 brain imaging phenotypes, we tested if brain imaging phenotypes had shared genetic architectures with non-imaging traits associated with the human brain. Of the 37 brain-related non-imaging traits with available large-scale GWAS summary statistics (Supplementary Table 12), 30 traits had at least one genome-wide significant locus containing an index variant of the pooled significant associations for brain imaging phenotypes. To test whether a trait-related locus has greater probability of colocalizing with brain imaging phenotypes than a genomic region of similar size (500 kb), which was generated by evenly segmenting the genome into chunks of 500 kb. The Fisher’s exact test (FDR-corrected q < 0.05) revealed that the associated loci of 19 traits had greater probability of colocalizing with brain imaging phenotypes than any evenly segmented genomic region. Then we used the resampling strategy to assess whether the number of colocalizations of trait-related loci with the index variants of brain imaging phenotypes is significantly higher than the number of colocalizations of these loci with the randomly generated 6,361 variants (1,000 sets). Each set of the created 6,361 variants was matched with the index variants of brain imaging phenotypes in allele frequency, gene proximity, and the number of LD proxies using the EUR and EAS from 1KGP as reference panels, respectively. We found that 16 traits showed larger number of colocalizations than random ( P < 0.05) in both ancestries and an additional trait was significant only in EAS. The 16 traits with shared genetic architectures with brain imaging phenotypes included cognitive performance, educational attainment, intelligence, neuroticism, risky behaviors, risk tolerance, insomnia symptom, drinks per week, smoking initiation, bipolar disorder, major depressive disorder, schizophrenia, any stroke, migraine, multiple sclerosis, and Parkinson’s disease (Fig. 5 a and Extended Data Fig. 4 ). To determine the specific loci shared by each pair of brain imaging phenotype and brain-related non-imaging trait, we only included the loci of the non-imaging trait whose index variants showed a high LD (r 2 > 0.8) with the index variants of the brain imaging phenotype and found 1,866 colocalizations based on this criterion (Fig. 5 b and Supplementary Table 13). For example, rs13107325 associated with the left putamen volume was also linked to five non-imaging traits including cognitive performance, intelligence, risky behaviors, drinks per week, and schizophrenia (Fig. 5 c). This missense mutation encodes SLC39A8 protein, which mediates the cellular uptake of zinc and manganese, two divalent metal cations that are important for development, tissue homeostasis and immunity 28 . Conclusions With the neuroimaging genetics data of 7,058 healthy Chinese Han participants from the CHIMGEN study and 33,224 EUR individuals from the UKBB project 15 , we conducted EAS-GWASs and trans-ancestral GWAS meta-analyses for 3,414 brain imaging phenotypes. The results deepen our understanding of the genetic architectures of brain imaging phenotypes by providing 496 novel associations that were discovered at P < 1.46 × 10 − 11 and replicated at P < 0.05 and highlight the importance of a more global representation of populations in neuroimaging genetics studies. The summary statistics of these GWASs based on Chinese Han participants are freely available on the CHIMGEN website ( http://chimgen.tmu.edu.cn/pheweb/ ), which provides public access to browse associations by variant, gene, or phenotype. This website was built with PheWeb ( https://github.com/statgen/pheweb/ ). By pooling genome-wide significant associations for brain imaging phenotypes in either single-ancestral or trans-ancestral GWASs, we identified 6,361 independent significant associations unevenly distributed in the genome and across subgroups of neuroimaging phenotypes and brain regions, which delineates a landscape of genetic effects on brain structure and function. These genetic associations with brain imaging phenotypes provide candidates for exploring causal genetic mechanisms for brain structure and function and identifying causal genome-brain-disorder pathways 29 . The functional enrichment results may improve our understanding of how genomic variations regulate the structural and functional properties of the human brain. A unique contribution of this study to the field of neuroimaging genetics is the differentiation of ancestry-specific associations from ancestry-shared associations. Consistent with prior studies on non-imaging traits 16 , 17 , brain imaging phenotypes also showed more ancestry-shared associations than ancestry-specific associations, indicating that most of the genetic associations for brain imaging phenotypes in one ancestral population can be extrapolated to other populations. However, we also provide reliable evidence for 43 ancestry-specific associations, which may account for the inter-ancestral difference in brain imaging phenotypes 10 . These ancestry-specific associations are also potential targets for studying inter-ancestral differences of other brain-related traits including brain disorders. Another valuable aspect of this study is the finding of colocalizations between brain imaging phenotypes and many brain-related non-imaging traits including cognition, personality, behavior, addiction, and neuropsychiatric disorders, which may improve our understanding of genetic mechanisms underlying the associations between brain imaging and non-imaging phenotypes. These results are useful for the development of diagnostic and therapeutic approaches of neuropsychiatric disorders 30 , especially valuable for mental disorders in which their diagnoses are based on a descriptive collection of behaviors without any objective test to stratify patients 31 . Genetic loci shared by brain imaging phenotypes and brain disorders are more worthy of biological validation and mechanistic studies, which can accelerate the discovery of novel biomarkers for diagnosis and new targets for treatment. Leveraging genetic instruments, Mendelian randomization can identify brain imaging phenotypes that are causally associated with mental disorders 29 , which are more reliable neuroimaging markers of mental disorders than those derived from the intergroup comparisons of neuroimaging data between patients and controls. Methods Participants EAS participants. All participants of East Asian ancestry (EAS) were recruited from the CHIMGEN study ( http://chimgen.tmu.edu.cn/ ) that has collected genomic, environmental, neuroimaging, and behavioral data from 7,306 healthy Chinese Han participants aged 18–30 years from 32 research centers located in 21 mainland cities of China from 2015 to 2019. The information and distribution of participants across centers is presented in Supplementary Table 14 and Extended Data Fig. 5 a. The CHIMGEN study was reviewed and approved by the Medical Research Ethics Committee of Tianjin Medical University General Hospital and was further reviewed and approved by corresponding local ethics committee of each research center, and written informed consent was obtained from each participant. The inclusion and exclusion criteria of the CHIMGEN participants are provided in Supplementary Table 15. The following criteria were further applied to filter the EAS participants by excluding: (a) participants without blood sample for genotyping (n = 111); (b) participants who failed to pass the quality control (QC) of genomic data (n = 32); and (c) participants who failed to pass the QC of magnetic resonance imaging (MRI) data [n = 105 for structural MRI (sMRI); n = 181 for diffusion MRI (dMRI); and n = 853 for resting-state functional MRI (rs-fMRI)] (Extended Data Fig. 6). The EAS participants scanned by the same type of MRI scanners (GE Discovery MR750) with the same parameters were defined as the discovery dataset to reduce the bias resulted from inconsistency in MRI data acquisition, and the remaining participants scanned by other scanners or parameters were defined as the replication dataset to replicate the discovered findings. In genome-wide association studies (GWASs), we finally included 7,058 EAS participants (5,025 for discovery and 2,033 for replication) for brain structural imaging phenotypes, 6,982 EAS participants (4,969 for discovery and 2,013 for replication) for brain diffusion imaging phenotypes, and 6,310 EAS participants (4,645 for discovery and 1,665 for replication) for brain functional imaging phenotypes. The demographic characteristics of the finally included EAS participants for GWASs for brain imaging phenotypes calculated from each MRI modality are listed in Supplementary Table 16, and the number of the included EAS participants from each research center for GWASs of brain imaging phenotypes obtained from each MRI modality is presented in Supplementary Table 14. EUR participants. All participants of European ancestry (EUR) were recruited from the UK Biobank (UKBB) study 4 , 15 , 32 since this study has collected similar genomic and neuroimaging (sMRI, dMRI, and rs-fMRI) data as the CHIMGEN study in more than 30,000 EUR participants. GWAS summary statistics of brain imaging phenotypes from two prior studies with different sample sizes 4 , 15 were used in this study with different purposes. The dataset with more brain imaging phenotypes (3,414 can be obtained from the CHIMGEN study) and more participants (33,224 EUR participants: 22,138 for discovery and 11,086 for replication) 15 was used in the main analyses to enhance statistical power. The dataset with 2,640 of the 3,414 brain imaging phenotypes obtained from 8,428 EUR participants 4 was used to reduce the potential bias resulting from the difference in sample size between EAS and EUR populations. Imaging data processing in CHIMGEN Imaging data acquisition. In the CHIMGEN study, brain imaging data were acquired with ten types of 3.0-Tesla MRI scanners and twelve sets of scanning parameters. The sMRI data were used to calculate brain imaging phenotypes that characterize the structural properties of the cerebrum, cerebellum, brain stem, and subcortical structures; the dMRI data were used to calculate brain imaging phenotypes that reflect diffusion properties of brain white matter tracts; and the rs-fMRI data were used to calculate brain imaging phenotypes that represent brain functional activity amplitude in and functional connectivity between resting-state networks (RSNs). Detailed scanning parameters for each MRI modality for different types of scanners are provided in Supplementary Tables 17–19, and the distribution of participants across types of scanners is presented in Extended Data Fig. 5 b. Imaging data QC before preprocessing. To obtain high-quality brain MRI data, a series of QC procedures were applied before and after the acquisition of brain MRI data in the CHIMGEN study. For example, we optimized scanning parameters for each MRI scanner before acquisition and identified and excluded MRI data with visible lesions, anatomical abnormalities, imaging artefacts, parameter inconsistency, and incomplete brain coverage immediately after acquisition. Imaging data preprocessing. To extract brain imaging phenotypes accurately and reliably, we developed a series of standardized pipelines to preprocess the multi-modal MRI data from the CHIMGEN study. These pipelines integrated the state-of-the-art preprocessing procedures and tools with the multi-cluster parallel computation. The total computing time was greatly reduced by using the Tianhe super-computer ( https://www.nscc-tj.cn ). The specific preprocessing pipeline for each type of the MRI data were as follows: sMRI data preprocessing for voxel-based morphometry (VBM). The CAT12 software ( http://dbm.neuro.uni-jena.de/cat ) was used to preprocess sMRI data for the VBM analyses. The VBM preprocessing steps included: (1) Bias correction: Image inhomogeneity caused by B1-field bias was corrected to segment the brain tissue accurately. (2) Segmentation: The bias-corrected structural MR images were segmented into gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF) using a model based on an adaptive Maximum A Posterior (MAP) technique 33 , which does not need a priori information about tissue probabilities. (3) Creating population-specific tissue templates: To improve the performance of image spatial normalization, the population-specific tissue probability templates for GM, WM, and CSF in the Montreal Neurological Institute (MNI) space were derived from 6,000 CHIMGEN participants using the Diffeomorphic Anatomical Registration Through Exponentiated Lie Algebra (DARTEL) algorithm 34 , which was implemented in Statistical Parametric Mapping 12 (SPM12) ( https://www.fil.ion.ucl.ac.uk/spm/software/spm12/ ). (4) Spatial normalization: The segmented GM images were spatially normalized to the population-specific GM template using the DARTEL algorithm and were resampled into a cubic voxel of 1.5 mm. Modulation was then performed on the normalized GM images to preserve the absolute GM volume (GMV). (5) Phenotype extraction (n = 143): From each participant, we extracted the total volumes of GM, WM, CSF, and GM + WM based on brain tissue segmentation and GMVs of 96 cerebral cortical subregions, 28 cerebellar subregions, and 15 subcortical nuclei and brain stem (Supplementary Table 20). sMRI data preprocessing for surface-based morphometry (SBM). FreeSurfer v6.0.0 ( http://surfer.nmr.mgh.harvard.edu/ ) was used to preprocess sMRI data for the SBM analyses with default settings. Specifically, the SBM preprocessing steps included: (1) Skull stripping: An automated skull-stripping was performed to separate the brain from non-brain tissues in structural MR images. Intensity normalization was applied before and after skull stripping to correct for the intensity non-uniformity due to variations in the sensitivity of reception coils and gradient-driven eddy currents. (2) Tissue segmentation: A series of tissue segmentation procedures were conducted based on intensity and neighbor constraints to generate the subcortical structures and the boundary between GM and WM. (3) Surface reconstruction: A two-dimensional tessellated mesh was constructed based on the WM-GM boundary to generate the WM surface, and the WM surface was extended outwards by tracking the GM intensity gradient to generate the pial surface. Topology correction was performed to repair topological defects. (4) Metric calculation: Surface-based metrics including the cortical thickness, surface area, and cortical volume were calculated based on the pial and WM surfaces. (5) Estimating spherical normalization parameters: Individual surfaces were then inflated into a spherical space and normalized to the fsaverage template to obtain the spherical normalization parameters. (6) Phenotype extraction (n = 1,035): Surface-based metrics for cortical regions defined by the surface atlases were extracted after converting them from standard to individual space using the inverse spherical normalization parameters. From each participant, we extracted the thickness of 306 cortical regions, the surface area of 370 cortical regions, the volume of 304 cortical regions and 55 subcortical regions (Supplementary Table 20). sMRI data preprocessing for subcortical segmentation. FMRIB’s Integrated Registration and Segmentation Tool (FIRST) ( https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FIRST ) was used to segment subcortical structures with default settings. The preprocessing steps included: (1) Intensity normalization: We corrected for the intensity non-uniformity due to variations in the sensitivity of reception coils and gradient-driven eddy currents. (2) Estimating spatial normalization parameters: A two-stage affine registration was performed to convert the intensity-normalized images of each participant to the MNI152 space to obtain spatial normalization parameters. (3) Segmentation: The inverse normalization parameters were used to convert shape models embedded in FIRST (provided by the Center for Morphometric Analysis) from standard to individual space where the segmentation was performed. Based on these converted models, FIRST searched for the most probable surface mesh of each subcortical structure according to the intensities from the input intensity-normalized images. Then mesh-based subcortical structures were converted to boundary corrected volumetric subcortical structures. (4) Phenotype extraction (n = 15): Based on the FIRST segmentation, we extracted the volumes of brain stem and 14 subcortical regions from each participant. dMRI data preprocessing. dMRI data were preprocessed using FMRIB Software Library (FSL, version 5.0.10; http://www.fmrib.ox.ac.uk/fsl ) with the following steps: (1) Skull stripping: The non-brain tissues were removed from the b = 0 images to generate a binary mask for the following tensor metric calculation. (2) Motion and distortion correction: The eddy_openmp program was used to evaluate and repair image displacement and signal dropout caused by head motion, and image distortion caused by eddy current. (3) Tensor metric calculation: The linear least square algorithm was used to estimate the diffusion tensor and to calculate diffusion metrics of each voxel from the tensor using the DTIFIT program. The obtained diffusion metrics included the three eigenvalues (L1, L2, and L3), mean diffusivity (MD), fractional anisotropy (FA), and mode of anisotropy (MO). (4) Estimating spatial normalization parameters: A two-step procedure was used to estimate the normalization parameters between individual diffusion and MNI standard space. Specifically, individual b = 0 images were aligned to structural images using the Boundary-Based Registration (BBR) algorithm. The obtained BBR transformation matrix was then concatenated with the DARTEL deformation field from individual to MNI space generated in the VBM preprocessing. The merged deformation field and its inverse deformation field were finally used in the following analyses. (5) Probabilistic fiber tracking: The BEDPOSTX program was used to estimate the diffusion orientation distribution based on a ball-stick model with the following parameters: maximum number of fibers per voxel = 3, burn-in period = 1,000, number of iterations = 1,250, and deconvolution model = sticks with a range of diffusivities. After converting the pre-defined seed, target, exclusion, and stop masks of AutoPtx ( https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/AutoPtx ) from standard to individual diffusion space using the inverse deformation field, the probabilistic fiber tracking was performed using the PROBTRACKX program with these converted masks in the individual diffusion space with the parameters of number of samples = 5,000 and angle threshold = 80 degree. (6) Metric normalization: The diffusion metrics were normalized into the MNI space using the above-mentioned merged deformation field and resampled into a cubic voxel of 2 mm. (7) Estimating diffusion metrics on white matter skeleton: We used a modified tract-based spatial statistics (TBSS) pipeline to create the white matter skeleton. In contrast to the standard TBSS pipeline 35 that directly aligns the individual FA images to the averaged FA template (FMRIB-58) in MNI space using the FNIRT program, we normalized individual FA images using the merged deformation field. Then a mean FA image of all individuals was created and “thinned” to generate a mean white matter skeleton representing the centers of white matter tracts common to all individuals. The aligned FA images of each participant were then projected onto the mean white matter skeleton by filling the mean skeleton with FA values from the nearest tract center, which was achieved by searching perpendicular to the local skeleton structure for maximal value. The obtained projection parameters were applied to other diffusion metric images (L1, L2, L3, MD, and MO) of the participant to estimate the diffusion properties of each voxel on the white matter skeleton. (8) Phenotype extraction (n = 450): From each participant, we extracted the L1, L2, L3, MD, FA, and MO of 75 white matter tracts (Supplementary Table 20). rs-fMRI data preprocessing . The rs-fMRI data preprocessing included the following steps: (1) Discarding unstable volumes: Since rs-fMRI data were acquired using three repetition times (TRs: 0.71, 0.8, and 2 seconds) and the functional images acquired at each TR were defined as a functional volume, the first functional volumes (≈ 10 seconds: 15, 13, 5 volumes for different TRs) were discarded to allow signal to reach equilibrium and to ensure the participants to adapt to scanning noise. (2) Slice timing correction: The remaining volumes were corrected for intra-volume temporal differences using sinc-interpolation. (3) Head motion correction: Inter-volume head motion was corrected by realigning each volume to the mean volume using a six-parameter rigid-body transformation. The frame-wise displacement (FD) was calculated by the Jenkinson approach 36 to index volume-to-volume changes in head position. When the FD of a volume was greater than 0.5 mm, this volume and its one previous volume and two subsequent volumes were defined as affected volumes. (4) Spatial normalization: After removing non-brain tissues from the head-motion-corrected functional images, the obtained functional images were co-registered to the structural images using the BBR method. Then all co-registered functional volumes were spatially normalized to the MNI space using the deformation field obtained from the VBM preprocessing and resampled to 3-mm isotropic voxels. (5) Noise reduction and bandpass filtering: Nuisance covariates including linear trend, Friston-24 head motion parameters, affected volumes, and WM and CSF signals were regressed out, and temporal bandpass filtering (0.01–0.08 Hz) was applied to reduce low-frequency drift and high-frequency noise. (6) Defining RSNs: RSNs were defined as the kept components from two group independent component analyses (group-ICA) conducted based on UKBB data 4 . We also kept 21 RSNs from the 25-component ICA and 55 RSNs from the 100-component ICA. (7) Back-reconstruct: A dual-regression method was used to back-reconstruct RSNs of each participant. (8) Defining functional metrices: After regressing out characteristic time courses of the non-RSN components, the characteristic time courses of RSNs were used to define two functional metrics. The functional activity amplitude of each RSN was defined as the standard deviation of fluctuations of the noise-removed characteristic time course of the RSN. The functional connectivity was defined as the temporal correlation of the noise-removed characteristic time courses between every two RSNs. (9) Phenotype extraction (n = 1,771): From each participant, we extracted 76 phenotypes to represent functional activity amplitude of each RSN and 1,695 phenotypes to reflect functional connectivity between every two RSNs (Supplementary Table 20). QC of the preprocessed MRI data. We also checked the preprocessed MRI data to find errors or imperfections emerged during imaging data preprocessing. If they were identified in a participant, we first tried to find reasons for imperfection and to re-run the pipeline after fixing them. If the newly obtained preprocessed MR images were still problematic, we had to exclude the participant from the following analyses for brain imaging phenotypes derived from the problematic imaging modality. The errors and imperfections mainly included bad tissue segmentation, imperfect spatial normalization, incorrect non-brain tissue removal, intensity normalization error, pial surface misplacement, topological defect, and fiber tracking error. The SPM12 was used to evaluate head motion in the rs-fMRI data. If the maximum displacement in any of the three orthogonal directions was more than 3.0 mm or a maximum rotation was greater than 3.0 degree, the rs-fMRI data of this participant would be excluded. We also excluded rs-fMRI data of participants with mean FD > 0.5 mm or with affected volumes more than one third of the total volumes. Extraction of brain imaging phenotypes. To perform trans-ancestral GWAS meta-analyses of brain imaging phenotypes, we generated 3,414 brain imaging phenotypes from EAS participants (CHIMGEN), all of which were also included in the previous GWASs in EUR participants (UKBB) 15 . These included 1,193 phenotypes of volume, surface area, and cortical thickness generated from sMRI data, 1,771 phenotypes of functional activity amplitude and functional connectivity obtained from rs-fMRI data, and 450 phenotypes of brain diffusion properties derived from dMRI data. A list of the included 3,414 brain imaging phenotypes is provided in Supplementary Table 20. Harmonization and normalization of brain imaging phenotypes. The use of different scanners and parameters may bring bias to the integrated analyses of brain imaging phenotypes obtained from multiple centers. Therefore, we applied ComBat harmonization to the 3,414 brain imaging phenotypes to remove variations of phenotypes resulting from scanner and parameter differences while preserving biologically relevant information 37 . Here, we first assessed the performance of ComBat harmonization in two subjects who traveled to and were scanned at 28 centers. For each brain imaging measure, the inter-scanner consistency was assessed by the correlations of this measure across brain regions between every two scanners in each subject. Brain imaging measures acquired from different MR scanners showed different degrees of inter-scanner inconsistencies before harmonization; however, the inter-scanner consistencies of these measures were greatly improved after ComBat harmonization (Extended Data Fig. 7a-d). Then, we calculated the distributions of each brain imaging phenotype across participants for each MR scanner before and after ComBat harmonization. We found that the distributions of brain imaging phenotypes became more similar across different MR scanners after harmonization (Extended Data Fig. 7e and 7f). The distributions of brain imaging phenotypes varied considerably even after harmonization, with a portion of phenotypes showing skewed distribution that would violate the assumption of the normal distribution of the phenotypic data when using the linear regression model to perform GWAS analyses. Thus, normal score transformation was applied to brain imaging phenotypes to make the data normally distributed and reduce undue influence of outliers. Specifically, the real values of each brain imaging phenotype were ranked from lowest to highest and these ranks were matched to and then replaced by equivalent ranks of random numbers generated from a normal distribution. Imaging data processing in UKBB The acquisition, preprocessing, and quality control of brain MRI data, and the extraction, harmonization, and normalization of brain imaging phenotypes have been described elsewhere 4 , 38 . The data preprocessing and metric calculation approaches for the 3,414 brain imaging phenotypes were similar between CHIMGEN and UKBB. Genetic data processing in CHIMGEN Blood collection, DNA extraction and genotyping. After blood sample collection, centrifugation and isolation were applied immediately in each research center to obtain plasma and buffy coat of each participant, which were then transported to the Tianjin Medical University General Hospital by a professional biomedical cold chain logistics company for centralized management and unified preprocessing. After standardized storage, the buffy coat was delivered to a sequencing company (Novogene Bioinformatics Technology, https://en.novogene.com/ ) for DNA extraction and genotyping. The CWE2100 Blood DNA Kit was used to extract DNA following the manufacturers’ specifications. In participants with qualified DNA samples defined as greater than 10 ng/µl of concentration and 260/280 between 1.8 and 2.2 of purity, the Asian Screening Array 750K (ASA-750K) specially designed for Asian populations was applied to 1ug DNA to capture genome-wide genetic variations. With one plate position for one sample, a 96-position plate can simultaneously deal with 96 DNA samples. To validate the reproducibility of genotyping, we deliberately included 86 blind duplicates (one duplicate for one plate) in the experiment. We then calculated the concordance rate of the genotyping results for every duplicate pair and found high concordance rates (ranging from 99.79–99.97%) (Extended Data Fig. 9a). Pre-imputation QC for genotyped data. Given the high homogeneity in racial identity (Chinese Han) in the CHIMGEN participants, a putative QC pipeline was used for the genotyped data with PLINKv2.0 39 ( http://zzz.bwh.harvard.edu/plink/ ). The QC procedures for CHIMGEN genetic data (7,195 participants genotyped at 743,722 variants) are presented in Extended Data Fig. 8, and the specific procedures were as follows: (1) Sex checking: We used genotyped data from the X chromosome to infer the sex of each participant and compared the result with the sex reported by the participant. We defined the inconsistency between the inferred and reported sex as sex mismatch, which is possibly caused by sample mishandling or DNA contamination. Since males only have one copy of X chromosome, they are expected to be homozygous for X markers outside the pseudo-autosomal region. Thus, the homozygosity rate of the X chromosome should be greater than 0.8 in males and smaller than 0.2 in females 40 . Based on these criteria, we found and excluded 2 participants with sex mismatch. (2) Sample-level missing rate: After excluding 17,555 duplicated variants from the 743,722 genotyped variants, we computed the genotype missing rate for each participant using the --miss command in PLINK in the remaining 726,167 variants. We found that 4 out of the 7,195 participants had a missing rate greater than 3% (Extended Data Fig. 9b), and these participants were excluded from the following analyses. (3) Variant-level missing rate: We also calculated genotype call rate for each of the 726,167 variants in the 7,191 participants. We found 21,612 variants with a missing rate > 5% (a genotype call rate < 95%), and these variants were excluded. (4) Minor allele frequency (MAF): We also computed MAF for each of the remaining 704,555 variants in the 7,191 participants. We found and excluded 142,828 variants with MAF < 0.001 and retained 561,727 variants for the next QC step (Extended Data Fig. 9d). (5) Hardy-Weinberg equilibrium (HWE): We estimated the deviation of each of the 561,727 variants from HWE and excluded 12,418 variants that were significantly deviated from HWE ( P < 10 − 6 ). The retained 549,309 variants were entered into the following QC procedures. (6) Heterozygosity rate: The extremely high heterozygosity rate is an indicator of poor DNA quality. Heterozygosity is defined by (N − O)/N, where N is the number of non-missing genotypes and O is the observed number of homozygous genotypes for a given participant. Based on the 549,309 variants, we calculated the heterozygosity rate for each of the remaining 7,191 participants using the --het command in PLINK. We excluded 15 participants with a heterozygosity greater than five times standard deviations from the mean (Extended Data Fig. 9b). (7) Related participants: A requirement of the population-based GWAS is that all included participants are unrelated. If duplicates and relatives are present, a bias will be introduced in the study because the genotypes within families are overrepresented 40 . Although we have paid attention to the requirement during the recruitment of participants, the property of multi-center and large-scale dataset made it unavoidable to include unintended duplicates or relatives. To identify duplicate and related individuals, we calculated identity by descent (IBD) for each pair of individuals based on independent variants. After removing the genomic regions with extended linkage disequilibrium (LD) entirely, the remaining genomic regions were pruned until no pair of variants within a given window (100 variants) was correlated (r 2 > 0.2). The remaining variants were defined as independent variants. Any pair of participants with an IBD > 0.1875 was considered as duplicate or related individuals. We found 11 duplicate or related pairs and then we excluded the one with higher sample-level missing rate from the following analyses. (8) Population structure: We used principal component analysis (PCA) to capture population structure of the CHIMGEN cohort, and the PCA resulted in 20 principal components (PCs). The PCA was conducted to identify individuals deviated from the Chinese Han population by projecting all participants onto the first two components of the four HapMap3 populations (CEU, CHB, JPT, YRI). We found one participant with extreme deviation from the CHIMGEN cohort (Extended Data Fig. 9e), and then this participant was excluded from the following analyses. In summary, the variant-level quality control excluded 17,555 duplicated variants, 21,612 variants with missing rate > 5%, 12,418 variants significantly deviated from HWE ( P < 10 − 6 ), and 142,828 variants with MAF < 0.001. Finally, 549,309 variants out of the 743,722 genotyped variants were included in genomic imputation. In the sample-level quality control, we excluded 2 sex mismatching participants, 11 duplicate or related participants, 4 participants with a genotype missing rate greater than 3%, and 15 participants with a heterozygosity greater than five times standard deviations from the mean (1 participant also with outlying population structure). Finally, 7,163 out of the 7,195 participants were included in the following analyses. To evaluate genotyping quality of the CHIMGEN participants, we calculated the across-variant correlation of allele frequencies in 524,924 overlapping variants between CHIMGEN and SG10K (another cohort of Asian population) 41 . We found a significant correlation (Spearman correlation: r = 0.96, P < 1.00 × 10 − 322 ) in allele frequencies between the two Asian cohorts (Extended Data Fig. 9c). This finding indicates high genotyping quality of the CHIMGEN participants, although the CHIMGEN and SG10K cohorts had different sample sizes (7,163 versus 4,810) and slightly different ancestral backgrounds (Chinese Han versus Asian populations) and used different genomic technologies (genotyping versus sequencing). Imputation and post-imputation QC in CHIMGEN The haplotypes were estimated based on the genotypes of the 549,309 qualified variants using SHAPEIT2 42 ( https://mathgen.stats.ox.ac.uk/genetics_software/shapeit/shapeit.html ) with default parameters. The phased autosomal variants were then imputed using the IMPUTE2 43 ( http://mathgen.stats.ox.ac.uk/impute/impute_v2.html ) in chunks of 5,000 kb with a combined reference panel that merged 1000 Genomes Project (1KGP, phase 3; n = 2,504) and SG10K (n = 4,471). We evaluated the imputation performance of the proposed scheme from the following two aspects: (1) Imputation accuracy among reference panels We compared the accuracy of imputation results obtained from three different reference panels. The first reference panel was the 1KGP panel that was downloaded from the website of IMPUTE2 ( https://mathgen.stats.ox.ac.uk/impute/1000GP_Phase3.html ). The second reference panel was derived from the phased 4,471 unrelated participants selected from the SG10K data of 4,810 participants. In the 4,810 SG10K participants, the unrelated participants were identified by calculating kinship coefficients ( 0.05) and relatively independent variants (pruned using PLINK with parameters: --indep-pairwise 1000 80 0.1 ). To identify the maximum number of unrelated individuals, we listed all related pairs and iteratively removed individual that appeared most frequently in the list until the list was empty. The third reference panel (1KGP + SG10K) was generated using the following three steps: (a) imputing 1KGP to SG10K; (b) imputing SG10K to 1KGP; and (c) merging the two imputed datasets to construct the 1KGP + SG10K panel using the IMPUTE2 program ( -merge_ref_panels ). To compare imputation accuracy among reference panels, we extracted 43,726 variants on chromosome 2 from 7,163 CHIMGEN participants. The genotype calls of 4,373 variants (1 out of every 10 variants sorted by position) were masked and saved for the evaluation of imputation accuracy. For each reference panel, we pre-phased the chromosome 2 using a reference-based phasing followed by imputation. Imputation error rate was estimated by comparing the imputed genotypes for the masked variants with their genotyped results. In addition, we counted imputed variants with information scores (INFO) ≥ 0.8 under four continuous MAF bins (0.005–0.01, 0.01–0.05, 0.05–0.2, and 0.2–0.5) for the results derived from each reference panel. We found that the 1KGP + SG10K panel had the best imputation performance over the other two panels (1KGP and SG10K). Specifically, the 1KGP + SG10K panel could impute the most high-quality variants (INFO ≥ 0.8) across all MAF bins (Extended Data Fig. 10a), while keeping the lowest imputation error rate (Extended Data Fig. 10b). (2) Imputation accuracy among genotyping arrays We also tested whether the ASA-750K specially designed for Asian populations could improve imputation performance in Chinese participants compared to other three arrays designed for populations with other ancestries, including the Illumina Global Screening Array, Affymetrix EUR, and Affymetrix Biobank that had comparable numbers of variants with the ASA-750K. Based on the genotype calls on chromosome 20 obtained from the high-coverage whole genome sequencing (WGS) data of 90 Chinese participants 45 , we extracted the genotypes of variants included in each array to mimic the genotyped data generated by that array. Then the extracted variants were used to impute other variants not included in the array with the 1KGP + SG10K panel. Since these 90 participants had been included in the 1KGP panel, we excluded these participants in the generation of the merged panel. We then compared the consistency of the imputed genotypes based on variants included in different arrays with the WGS genotypes on chromosome 20. The ASA-750K showed the best imputation accuracy for the CHIMGEN genetic data among the four arrays (Extended Data Fig. 10c). Finally, the 549,309 variants were imputed to 111,370,847 autosomal variants with the 1KGP + SG10K panel. Extended Data Fig. 10d shows the distribution of INFO on all markers in the imputed dataset. To avoid false positive signals in the association analyses, we filtered variants with MAF < 0.01 and INFO < 0.9, and finally kept 6,830,145 autosomal variants in the GWASs. The distributions of MAF and INFO of the finally included 6,830,145 autosomal variants are presented in Extended Data Fig. 10e. Genetic data processing and imputation in UKBB The detailed processing and imputation steps for the genetic data of UKBB were provided in a previous study 32 . Covariates for EAS-GWASs in CHIMGEN Since a variety of confounding factors might mask or bias the effects of genetic variants on brain imaging phenotypes, we controlled for a series of covariates in EAS-GWASs to reduce the risk of reporting false positive associations. We designed covariates according to the previous GWASs for brain imaging phenotypes 6 , 7 , including age, sex, age × sex, age × age, and top three PCs for EAS-GWASs for all brain imaging phenotypes. Only the top three PCs were selected as covariates because they could account for the main variance of population stratification in this highly homogeneous population (CHIMGEN) (Extended Data Fig. 11). The total intracranial volume (TIV) was also included in GWASs for regional brain structural imaging phenotypes (such as cortical volume, surface area, and cortical thickness), and mean FD (measuring head motion) was included in GWASs for brain functional imaging phenotypes (functional activity amplitude and functional connectivity). All included covariates were transformed into Z-scores and missing values were set to zero in the transformed data. The covariates used in EUR-GWASs in UKBB can be found in previous studies 4 , 15 . EAS-GWASs for brain imaging phenotypes in CHIMGEN For the yielded 6,830,145 autosomal variants from CHIMGEN, an additive model was applied to investigate the association between the dosage of each variant and each brain imaging phenotype using BGENEv1.2 4 ( https://jmarchini.org/bgenie/ ), which was specially designed to deal with GWASs of many phenotypes performed simultaneously. GWASs were performed for 3,414 brain imaging phenotypes based on up to 7,058 EAS participants, including 1,193 brain structural imaging phenotypes (5,025 participants for discovery and 2,033 participants for replication), 1,771 brain functional imaging phenotypes (4,645 participants for discovery and 1,665 participants for replication), and 450 brain diffusion imaging phenotypes (4,969 participants for discovery and 2,013 participants for replication), while controlling for the above-mentioned covariates. In the GWAS for each brain imaging phenotype, the independent variant-phenotype associations were identified by the following steps: (a) all variants with P < 5 × 10 − 8 in the discovery sample were used to create a list of variants; (b) a locus of 500 kb centered at the most significant variant (index variant) in the list was generated and all variants within the locus were removed from the list; (c) the remaining variants formed a new list of variants and then the step (b) was repeated; (d) the iterative process stopped until the list was empty; and (e) the iterative process would generate several loci with independent index variants, of which the overlapping loci were merged into an independent locus indexed by the most significant variant of these loci. In this way, we identified all independent variant-phenotype associations in the discovery sample of each GWAS. We also repeated the analyses with a threshold of P < 1.46 × 10 − 11 in the discovery samples to additionally correct for the 3,414 brain imaging phenotypes. In the replication samples, we reported the associations that were confirmed at uncorrected P < 0.05, false positive rate corrected (FDR-corrected) q < 0.05, and Bonferroni-corrected P < 0.05. Trans-ancestral GWAS meta-analyses for brain imaging phenotypes Trans-ancestral GWAS meta-analyses can boost the power to detect novel genetic associations when the underlying causal variants are shared between ancestries. For the 5,950,889 autosomal variants included in the CHIMGEN and UKBB datasets, a fixed-effect model embedded in the METASOFT tool 46 ( http://genetics.cs.ucla.edu/meta/ ) was used to perform trans-ancestral GWAS meta-analyses between EAS and EUR for 3,414 brain imaging phenotypes that were generated in similar ways from CHIMGEN and UKBB. GWAS summary statistics of the discovery stages of CHIMGEN (5,025 participants) and UKBB (22,138 participants) were used to conduct trans-ancestral GWAS meta-analyses in the discovery stage, while GWAS summary statistics of the replication stages of CHIMGEN (2,033 participants) and UKBB (11,086 participants) were used to perform trans-ancestral GWAS meta-analyses in the replication stage. The same approach as the EAS-GWASs was used to identify independent variant-phenotype associations in the trans-ancestral GWAS meta-analyses with different statistical thresholds and multiple testing correction methods in the discovery ( P < 5 × 10 − 8 and P < 1.46 × 10 − 11 ) and replication (uncorrected, FDR-corrected, and Bonferroni-corrected P < 0.05) samples. Identifying novel associations from EAS-GWASs and trans-ancestral GWASs To identify novel variant-phenotype associations from EAS-GWASs and trans-ancestral GWASs, we defined the known independent significant associations as those with P < 1.46 × 10 − 11 in the discovery stage (22,138 participants) and P < 0.05 in the replication stage (11,086 participants) in the currently largest EUR-GWASs for the 3,414 brain imaging phenotypes 15 . With the same thresholds ( P < 1.46 × 10 − 11 for discovery and P < 0.05 for replication), we created the lists of independent variant-phenotype associations for EAS-GWASs and trans-ancestral GWASs. In EAS-GWASs, a novel variant-phenotype association was defined as the corresponding locus of the variant that did not overlap with any known loci for the same brain imaging phenotype in EUR-GWASs. In trans-ancestral GWAS meta-analyses, a novel variant-phenotype association was defined as the corresponding locus of the variant that did not overlap with any loci of the same brain imaging phenotype in either EUR-GWASs or EAS-GWASs. Pooled genome-wide significant associations We pooled all genome-wide significant associations (discovered at P < 5 × 10 − 8 and replicated at P < 0.05) identified in any of the EAS-GWASs, EUR-GWASs, or trans-ancestral GWASs for each of the 3,414 brain imaging phenotypes. Specifically, for each phenotype, the overlapping associated loci among these three types of GWASs were merged into an independent locus indexed by the most significant variant of these loci. Enrichment analysis for the pooled genome-wide significant associations PANTHER ( http://pantherdb.org/ ) was used to conduct functional enrichment analysis. We first assigned the index variant of each pooled genome-wide significant association to a protein-coding gene (gencode.v38lift37.annotation.gtf.gz) if the variant located within 10 kb around the gene. Then the obtained genes were used to test functional enrichment in pathways derived from Reactome Pathway Database using all protein-coding genes (n = 20,589) as the background. Significance of functional enrichment was calculated using Fisher’s exact test with an FDR-corrected q < 0.05. Ancestry-shared and ancestry-specific associations between EAS and EUR The Cochran’s Q-test (CQ-test) was applied to examine the effect size differences in variant-phenotype associations between EAS and EUR. The pooled independent significant associations (discovered at P < 5 × 10 − 8 and replicated at P < 0.05) whose index variants studied in both EAS- and EUR-GWASs were included in the heterogeneous assessment with a discovery-replication scheme. In the discovery stage, the CQ-test was conducted based on GWAS summary statistics from the discovery samples of CHIMGEN (n = 5,025) and UKBB (n = 22,138). In the replication stage, the CQ-test was performed based on GWAS summary statistics from the replication samples of CHIMGEN (n = 2,033) and UKBB (n = 11,086). Based on the heterogeneous assessment, the variant-phenotype associations were divided into three categories of ancestry-shared, ancestry-specific, and un-classified associations. The ancestry-shared associations were defined as those variant-phenotype associations with P ≥ 0.05 in both discovery and replication CQ-tests. The ancestry-specific associations were defined as those variant-phenotype associations with Pc < 0.05 (Bonferroni correction for the number of the included associations) in the discovery CQ-tests and uncorrected, FDR-corrected, and Bonferroni-corrected P < 0.05 in the replication CQ-tests. Since the unbalanced sample sizes between EAS (n = 5,025 for discovery and n = 2,033 for replication) and EUR (n = 22,138 for discovery and n = 11,086 for replication) may bring bias to the heterogeneous assessment, we also validate the identified ancestry-shared and ancestry-specific associations in the EAS (n = 7,058 from CHIMGEN) and EUR (n = 8,428 from UKBB) 4 populations with comparable sample size. Because only 2,640 of the 3,414 brain imaging phenotypes were included in the EUR-GWASs 4 , we filtered out the identified ancestry-shared and ancestry-specific associations that were absent in the EUR-GWASs 4 . Therefore, the CQ-tests were only performed for the remaining ancestry-shared associations and ancestry-specific associations. The ancestry-shared associations were deemed to be validated when P ≥ 0.05. We reported the ancestry-specific associations that were validated at uncorrected, FDR-corrected, and Bonferroni-corrected P < 0.05. Genetic colocalizations between brain imaging phenotypes and brain-related non-imaging traits We tested whether brain imaging phenotypes had shared genetic architectures with various brain-related non-imaging traits. We only included non-imaging traits associated with brain structure and function and with available GWAS summary statistics. If more than one GWASs were available for a trait, we only included the GWAS summary statistics with the largest number of independent associations. For non-imaging traits with available GWAS summary statistics for both EUR and EAS populations, we performed trans-ancestral GWAS meta-analyses with fixed-effect model to integrate the results. The finally included 37 non-imaging traits consisted of 3 traits for cognition (cognitive performance, educational attainment, and intelligence); 13 for personality and behavior (aggressive behavior, antisocial behavior, depressive symptoms, subjective well-being, agreeableness, conscientiousness, openness, extraversion, neuroticism, risky behaviors, risk tolerance, insomnia symptoms and isolation); 2 for addiction (drinks per week and smoking initiation); 11 for psychiatric disorders (anorexia nervosa, anxiety disorder, attention deficit hyperactivity disorder, autism spectrum disorder, bipolar disorder, major depressive disorder, obsessive compulsive disorder, panic disorder, post-traumatic stress disorder, schizophrenia, and Tourette syndrome); and 8 for neurological disorders (Alzheimer’s disease, amyotrophic lateral sclerosis, any stroke, focal epilepsy, genetic generalized epilepsy, migraine, multiple sclerosis, and Parkinson’s disease) (Supplementary Table 12). We generated independent significant loci ( P < 5 × 10 − 8 ) for each non-imaging trait using an iterative process including the following procedures: (a) identifying the most significant variant; (b) grouping all variants within 500 kb centered at the variant into a locus; and (c) merging overlapping loci. Here, we used three continuous steps to identify shared genetic architectures between brain-related non-imaging traits and brain imaging phenotypes. In step 1, we screened non-imaging traits with at least one locus containing an index variant of the pooled genome-wide significant associations for brain imaging phenotypes ( P < 5 × 10 − 8 for discovery and P < 0.05 for replication). In step 2, we split the autosome into 11,508 chunks with 500kb (similar with the mean size of the significant loci for the non-imaging traits) and defined the probability of a chunk that contained at least one index variant for brain imaging phenotypes as the reference probability of non-significance. For each non-imaging trait survived in step 1, Fisher’s exact test was performed to test whether the colocalization probability of the trait with brain imaging phenotypes was significantly higher than the reference probability. In step 3, we used vSampler 47 to randomly generate 1000 sets of variants with matched allele frequency, gene proximity, and number of LD proxies with the index variants of brain imaging phenotypes using the EUR and EAS from 1KGP as reference panels, respectively. For each non-imaging trait also survived in step 2, we tested whether the number of colocalized trait-related loci with the index variants of brain imaging phenotypes was significantly higher than the number of colocalized trait-related loci with randomly generated variants using the resampling strategy (1,000 sets, P < 0.05). The non-imaging traits survived in the three continuous tests were deemed to have shared genetic architectures with brain imaging phenotypes. To determine the specific loci shared by each pair of imaging and non-imaging traits, we only included the loci of the non-imaging trait whose index variants showed a high LD (r2 > 0.8) with the index variants of the brain imaging phenotype. Declarations Acknowledgements We are grateful to participants and researchers of CHIMGEN, who generously donated their time to make this resource available. We acknowledge funding from the National Key Research and Development Program of China (2018YFC1314300) to Chunshui Yu, and the National Natural Science Foundation of China (82030053, 81425013) to Chunshui Yu. We thank SG10K Consortium for collecting and sharing the high-coverage whole genome sequencing data of Asian Populations. We acknowledge UK Biobank for providing GWAS summary statistics of brain imaging phenotypes. For the genetic colocalization analyses, we used summary statistical data from several GWASs of brain-related non-imaging traits. We thank groups [BioBank Japan (BBJ); the Complex Traits Genetics laboratory (CTGlab); the Early Genetics and Lifecourse Epidemiology consortium (EAGLE); Genetics of Personality Consortium (GPC); the GWAS and Sequencing Consortium of Alcohol and Nicotine use (GSCAN); the International Headache Genetics Consortium (IHGC); the International League Against Epilepsy (ILAE); International Multiple Sclerosis Genetics Consortium (IMSGC); METASTROKE collaboration, the Psychiatric Genomics Consortium (PGC); the Social Science Genetic Association Consortium (SSGAC); UK Biobank (UKBB)], authors (Jacqueline M Lane, Jia Nee Foo, Mike A Nalls and Wouter van Rheenen) for making these data publicly available and all the participants and researchers in these studies. Author contributions Chunshui Yu and Jilian Fu designed the study and wrote the manuscript. Jilian Fu, Jianhua Wang and Quan Zhang analyzed the data. Chunshui Yu, Mulin Jun Li and Jingliang Cheng supervised this work. Jilian Fu, Quan Zhang, Jianhua Wang, Meiyun Wang, Bing Zhang, Wenzhen Zhu, Shijun Qiu, Zuojun Geng, Guangbin Cui, Yongqiang Yu, Weihua Liao, Hui Zhang, Bo Gao, Xiaojun Xu, Tong Han, Zhenwei Yao, Wen Qin, Feng Liu, Meng Liang, Sijia Wang, Qiang Xu, Jiayuan Xu, Peng Zhang, Wei Li, Dapeng Shi, Caihong Wang, Su Lui, Zhihan Yan, Feng Chen, Jing Zhang, Jiance Li, Wen Shen, Yanwei Miao, Dawei Wang, Junfang Xian, Jia-Hong Gao, Xiaochu Zhang, Kai Xu, Xi-Nian Zuo, Longjiang Zhang, Zhaoxiang Ye, Jingliang Cheng, Mulin Jun Li and Chunshui Yu acquired the data. All authors critically reviewed the manuscript. Competing interests The authors declare no competing interests. Code availability All the software and code used in this study are publicly available BGENEv1.2 (https://jmarchini.org/bgenie/) CAT12 (http://dbm.neuro.uni-jena.de/cat) ComBat harmonization (https://github.com/precision-medicine-um/ComBatHarmonization) FMRIB’s Integrated Registration and Segmentation Tool (FIRST) (https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FIRST) FMRIB Software Library (FSL, version 5.0.10; http://www.fmrib.ox.ac.uk/fsl) FreeSurfer v6.0.0 (http://surfer.nmr.mgh.harvard.edu/) IMPUTE2 (http://mathgen.stats.ox.ac.uk/impute/impute_v2.html) KING (https://www.kingrelatedness.com/) METASOFT (http://genetics.cs.ucla.edu/meta/) PANTHER (http://pantherdb.org/) PheWeb (https://github.com/statgen/pheweb/) PLINKv2.0 (http://zzz.bwh.harvard.edu/plink/) SHAPEIT2 (https://mathgen.stats.ox.ac.uk/genetics_software/shapeit/shapeit.html) SPM12 ( https://www.fil.ion.ucl.ac.uk/spm/software/spm12/ ) Data availability All GWAS results based on Chinese Han participants are available on the website of CHIMGEN (http://chimgen.tmu.edu.cn/pheweb/), which allows users to browse associations by variants, genes, and brain imaging phenotypes. All GWAS results based on UK Biobank participants are available on the website of Oxford Brain Imaging Genetics (BIG) web browser (http://big.stats.ox.ac.uk/). 1000 Genomes Project reference panel can be found on https://mathgen.stats.ox.ac.uk/impute/1000GP_Phase3.html. And the high-coverage whole genome sequencing data of SG10K is available on https://ega-archive.org with accession number of EGAS00001003875. All authors of CHIMGEN Consortium Department of Radiology and Tianjin Key Laboratory of Functional Imaging, Tianjin Medical University General Hospital Chunshui Yu, Quan Zhang, Wen Qin, Feng Liu, Junping Wang, Qiang Xu, Jiayuan Xu, Xue Zhang, Xinjun Suo, Jilian Fu, Congcong Yuan, Yuan Ji, Hui Xue, Tianying Gao, Junpeng Liu, Yanjun Li, Xi Guo, Lixue Xu, Jiajia Zhu, Huaigui Liu, Fangshi Zhao, Jie Sun, Yongjie Xu, Huanhuan Cai, Jie Tang, Yaodan Zhang, Yongqin Xiong, Xianting Sun, Nannan Pan, Xue Zhang (Junior), Jiayang Yang, Nana Liu, Ya Wen, Dan Zhu, Bingjie Wu, Wenshuang Zhu, Qingqing Diao, Yujuan Cao, Bingbing Yang, Lining Guo, Yingying Xie, Jiahui Lin, Zhimin Li, Yan Zhang, Kaizhong Xue, Zirui Wang, Junlin Shen School of Medical Imaging, Tianjin Medical University Meng Liang, Xuejun Zhang, Hao Ding, Qian Su, Sijia Wang Department of Bioinformatics, The Province and Ministry Co-sponsored Collaborative Innovation Center for Medical Epigenetics, School of Basic Medical Sciences, Tianjin Medical University Mulin Jun Li, Shijie Zhang, Jianhua Wang Department of Radiology, Tianjin Medical University Cancer Institute and Hospital Zhaoxiang Ye, Peng Zhang, Wei Li Department of Radiology, Henan Provincial People’s Hospital & Zhengzhou University People’s Hospital Meiyun Wang, Dapeng Shi, Lun Ma, Yan Bai, Min Guan, Wei Wei Department of Magnetic Resonance Imaging, The First Affiliated Hospital of Zhengzhou University Jingliang Cheng, Caihong Wang, Peifang Miao, Fuhong Duan, Yafei Guo, Weijian Wang Department of Medical Imaging, Jinling Hospital, Medical School of Nanjing University Longjiang Zhang, Lijuan Zheng, Li Lin, Yunfei Wang, Han Zhang, Xinyuan Zhang Department of Radiology, Drum Tower Hospital, Medical School of Nanjing University Bing Zhang, Zhao Qing, Sichu Wu, Junxia Wang, Yi Sun, Yang He Institute of Psychology, Chinese Academy of Sciences Xi-Nian Zuo, Zhe Zhang, Yin-Shan Wang, Quan Zhou Department of Radiology, Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology Wenzhen Zhu, Tian Tian Department of Medical Imaging, The First Affiliated Hospital of Guangzhou University of Traditional Chinese Medicine Shijun Qiu, Yi Liang, Yujie Liu, Hui Zeng, Jingxian Chen Department of Radiology, The Affiliated Hospital of Xuzhou Medical University Kai Xu, Haitao Ge, Peng Xu, Cailuan Lu, Chen Wu, Xiaoying Yang Department of Medical Imaging, The Second Hospital of Hebei Medical University Zuojun Geng, Yuzhao Wang, Yankai Wu, Xuran Feng, Ling Li, Duo Gao Division of Life Science and Medicine, University of Science & Technology of China Xiaochu Zhang, Rujing Zha, Ying Li, Lizhuang Yang, Ying Chen, Ling Zuo Center for MRI Research, Academy for Advanced Interdisciplinary Studies, Peking University Jia-Hong Gao, Jianqiao Ge, Guoyuan Yang Functional and Molecular Imaging Key Lab of Shaanxi Province & Department of Radiology, Tangdu Hospital, Air Force Medical University Guangbin Cui, Wen Wang, Linfeng Yan, Yang Yang, Jin Zhang Department of Radiology, Beijing Tongren Hospital, Capital Medical University Junfang Xian, Qian Wang, Xiaoxia Qu, Ying Wang Department of Radiology, Characteristic Medical Center of Chinese People’s Armed Police Force Quan Zhang, Fei Yuan Department of Radiology, Qilu Hospital of Shandong University Dawei Wang, Li Hu, Jizhen Li Department of Radiology, The First Affiliated Hospital of Dalian Medical University Yanwei Miao, Weiwei Wang, Yujing Zhou Department of Radiology, Tianjin First Center Hospital Wen Shen, Miaomiao Long, Lihua Liu Department of Radiology, The First Affiliated Hospital of Anhui Medical University Yongqiang Yu, Xiaohu Li, Xiaoshu Li Department of Radiology, The First Affiliated Hospital of Wenzhou Medical University Jiance Li, Yunjun Yang, Nengzhi Xia Department of Radiology, Xiangya Hospital, Central South University Weihua Liao, Shuai Yang, Youming Zhang Department of Magnetic Resonance, Lanzhou University Second Hospital Jing Zhang, Guangyao Liu, Laiyang Ma Department of Radiology, The First Hospital of Shanxi Medical University Hui Zhang, Xiaochun Wang, Ying Lei Department of Radiology, Yantai Yuhuangding Hospital Bo Gao, Gang Zhang, Kang Yuan Department of Radiology, The Second Affiliated Hospital of Zhejiang University, School of Medicine Xiaojun Xu, Jingjing Xu, Xiaojun Guan Department of Radiology, Hainan General Hospital Feng Chen, Yuankai Lin, Huijuan Chen Department of Radiology, The Second Affiliated Hospital and Yuying Children’s Hospital of Wenzhou Medical University Zhihan Yan, Yuchuan Fu, Yi Lu Department of Radiology, Tianjin Huanhu Hospital Tong Han, Jun Guo, Hao Lu Department of Radiology, Huashan Hospital, Fudan University Zhenwei Yao, Yue Wu Department of Radiology, the Center for Medical Imaging, West China Hospital of Sichuan University Su Lui References Deco, G. & Kringelbach, M. L. Great Expectations: Using Whole-Brain Computational Connectomics for Understanding Neuropsychiatric Disorders. Neuron 84 , 892–905 (2014). Jansen, A. G., Mous, S. E., White, T., Posthuma, D. & Polderman, T. J. C. What Twin Studies Tell Us About the Heritability of Brain Development, Morphology, and Function: A Review. Neuropsychol Rev 25 , 27–46 (2015). Makowski, C. et al. Discovery of genomic loci of the human cerebral cortex using genetically informed brain atlases. Science 375 , 522–528 (2022). Elliott, L. T. et al. Genome-wide association studies of brain imaging phenotypes in UK Biobank. Nature 562 , 210–216 (2018). Zhao, B. X. et al. Common genetic variation influencing human white matter microstructure. Science 372 , 1304–1316 (2021). Hibar, D. P. et al. Common genetic variants influence human subcortical brain structures. Nature 520 , 224–229 (2015). Grasby, K. L. et al. The genetic architecture of the human cerebral cortex. Science 367 (2020). Kang, D. W. et al. Differences in cortical structure between cognitively normal East Asian and Caucasian older adults: a surface-based morphometry study. Scientific reports 10 (2020). Tang, Y. C. et al. Brain structure differences between Chinese and Caucasian cohorts: A comprehensive morphometry study. Human brain mapping 39 , 2147–2155 (2018). Tang, Y. C. et al. The construction of a Chinese MRI brain atlas: A morphometric comparison study between Chinese and Caucasian cohorts. NeuroImage 51 , 33–41 (2010). Langer-Gould, A., Brara, S. M., Beaber, B. E. & Zhang, J. L. Incidence of multiple sclerosis in multiple racial and ethnic groups. Neurology 80 , 1734–1739 (2013). Papp, V. et al. Worldwide Incidence and Prevalence of Neuromyelitis Optica A Systematic Review. Neurology 96 , 59–77 (2021). Meeker, K. L. et al. Socioeconomic Status Mediates Racial Differences Seen Using the AT(N) Framework. Annals of neurology 89 , 254–265 (2021). Peterson, R. E. et al. Genome-wide Association Studies in Ancestrally Diverse Populations: Opportunities, Methods, Pitfalls, and Recommendations. Cell 179 , 589–603 (2019). Smith, S. M. et al. An expanded set of genome-wide association studies of brain imaging phenotypes in UK Biobank. Nature neuroscience 24 , 737–745 (2021). Graham, S. E. et al. The power of genetic diversity in genome-wide association studies of lipids. Nature (2021). Chen, M. H. et al. Trans-ethnic and Ancestry-Specific Blood-Cell Genetics in 746,667 Individuals from 5 Global Populations. Cell 182 , 1198–1213 e1114 (2020). Xu, Q. et al. CHIMGEN: a Chinese imaging genetics cohort to enhance cross-ethnic and cross-geographic brain research. Molecular psychiatry 25 , 517–529 (2020). Mela, A. & Goldman, J. E. The Tetraspanin KAI1/CD82 Is Expressed by Late-Lineage Oligodendrocyte Precursors and May Function to Restrict Precursor Migration and Promote Oligodendrocyte Differentiation and Myelination. Journal of Neuroscience 29 , 11172–11181 (2009). Zhao, Y. et al. CD82-TRPM7-Numb signaling mediates age-related cognitive impairment. Geroscience 42 , 595–611 (2020). Suzuki, T. et al. Genome-wide association study of epilepsy in a Japanese population identified an associated region at chromosome 12q24. Epilepsia 62 , 1391–1400 (2021). Hautakangas, H. et al. Genome-wide analysis of 102,084 migraine cases identifies 123 risk loci and subtype-specific risk alleles. Nature genetics 54 , 152–160 (2022). Hong, S. E. et al. Autosomal recessive lissencephaly with cerebellar hypoplasia is associated with human RELN mutations. Nature genetics 26 , 93–96 (2000). Mao, D. X. et al. De novo EIF2AK1 and EIF2AK2 Variants Are Associated with Developmental Delay, Leukoencephalopathy, and Neurologic Decompensation. American journal of human genetics 106 , 570–583 (2020). Chen, H. R., Juan, H. C., Wong, Y. H., Tsai, J. W. & Fann, M. J. Cdk12 Regulates Neurogenesis and Late-Arising Neuronal Migration in the Developing Cerebral Cortex. Cerebral cortex 27 , 2289–2302 (2017). Greenbaum, L. & Lerer, B. Differential contribution of genetic variation in multiple brain nicotinic cholinergic receptors to nicotine dependence: recent progress and emerging open questions. Molecular psychiatry 14 , 912–945 (2009). Wight, T. N. Versican: a versatile extracellular matrix proteoglycan in cell biology. Current opinion in cell biology 14 , 617–623 (2002). Zang, Z. S., Xu, Y. M. & Lau, A. T. Y. Molecular and pathophysiological aspects of metal ion uptake by the zinc transporter ZIP8 (SLC39A8). Toxicol Res-Uk 5 , 987–1002 (2016). Song, W. C., Qian, W., Wang, W. D., Yu, S. Y. & Lin, G. N. Mendelian randomization studies of brain MRI yield insights into the pathogenesis of neuropsychiatric disorders. BMC genomics 22 (2021). Buch, A. M. & Liston, C. Dissecting diagnostic heterogeneity in depression by integrating neuroimaging and genetics. Neuropsychopharmacology: official publication of the American College of Neuropsychopharmacology 46 , 156–175 (2021). Abi-Dargham, A. & Horga, G. The search for imaging biomarkers in psychiatric disorders. Nat Med 22 , 1248–1255 (2016). Bycroft, C. et al. The UK Biobank resource with deep phenotyping and genomic data. Nature 562 , 203–209 (2018). Rajapakse, J. C., Giedd, J. N. & Rapoport, J. L. Statistical approach to segmentation of single-channel cerebral MR images. IEEE transactions on medical imaging 16 , 176–186 (1997). Ashburner, J. A fast diffeomorphic image registration algorithm. NeuroImage 38 , 95–113 (2007). Smith, S. M. et al. Tract-based spatial statistics: Voxelwise analysis of multi-subject diffusion data. NeuroImage 31 , 1487–1505 (2006). Jenkinson, M., Bannister, P., Brady, M. & Smith, S. Improved optimization for the robust and accurate linear registration and motion correction of brain images. NeuroImage 17 , 825–841 (2002). Fortin, J. P. et al. Harmonization of multi-site diffusion tensor imaging data. NeuroImage 161 , 149–170 (2017). Alfaro-Almagro, F. et al. Image processing and Quality Control for the first 10,000 brain imaging datasets from UK Biobank. NeuroImage 166 , 400–424 (2018). Purcell, S. et al. PLINK: a tool set for whole-genome association and population-based linkage analyses. American journal of human genetics 81 , 559–575 (2007). Anderson, C. A. et al. Data quality control in genetic case-control association studies. Nat Protoc 5 , 1564–1573 (2010). Wu, D. et al. Large-Scale Whole-Genome Sequencing of Three Diverse Asian Populations in Singapore. Cell 179 , 736–749 e715 (2019). Delaneau, O., Zagury, J. F. & Marchini, J. Improved whole-chromosome phasing for disease and population genetic studies. Nature methods 10 , 5–6 (2013). Howie, B., Fuchsberger, C., Stephens, M., Marchini, J. & Abecasis, G. R. Fast and accurate genotype imputation in genome-wide association studies through pre-phasing. Nature genetics 44 , 955–959 (2012). Manichaikul, A. et al. Robust relationship inference in genome-wide association studies. Bioinformatics 26 , 2867–2873 (2010). Lan, T. M. et al. Deep whole-genome sequencing of 90 Han Chinese genomes. GigaScience 6 (2017). Han, B. & Eskin, E. Random-Effects Model Aimed at Discovering Associations in Meta-Analysis of Genome-wide Association Studies. American journal of human genetics 88 , 586–598 (2011). Huang, D. D. et al. vSampler: fast and annotation-based matched variant sampling tool. Bioinformatics 37 , 1915–1917 (2021). Additional Declarations There is NO Competing Interest. Supplementary Files supplementarytables.xlsx Dataset 1 ExtendedDataFig2.tif Associations of rs67827860 with brain diffusion imaging phenotypes in CHIMGEN. The rs67827860 shows significant associations with many brain diffusion imaging phenotypes. Colors of points represent the subgroups of phenotypes. Dashed line means a cutoff at Bonferroni-corrected threshold of P < 1.46 × 10−5. FA, fractional anisotropy; L1, maximal eigenvalue; L2, medial eigenvalue; L3, minimal eigenvalues; MD, mean diffusivity; MO, mode of anisotropy. ExtendedDataFig3.tif Spatial distribution of pooled independent significant associations in each brain atlas or parcellation. a, Diedrichsen cerebellar atlas (SUIT); b, Harvard-Oxford subcortical atlas; c,Harvard-Oxford cortical atlas; d, Subcortical structures from FIRST; e, Subcortical structures from aseg; f, BA exvivo parcellation; g, Desikan-Killiany (DK) parcellation; h, Desikan-Killiany-Tourville (DKT) parcellation; i, Parcellation based on the pial surface using Desikan-Killiany parcellation (pial); j, JHU fibers; k, UKBB 100-component group-ICA. In a-j, colors represent the numbers of significant associations for each brain region. In k, circle colors reflect the numbers of significant associations for functional activity amplitude of each brain RSN and a link indicates the existence of a significant association for functional connectivity between every two brain RSNs. ExtendedDataFig4.tif Distinguishing true genetic sharing between brain-related non-imaging traits and brain imaging phenotypes from colocalizations by chance. a, b, For each brain-related non-imaging trait, a random distribution of colocalizations (blue) is generated by the number of trait-related loci colocalized with each of 1,000 sets of randomly created variants matching with the index variants of 6,361 pooled independent genome-wide significant associations ( P < 5 × 10−8 for discovery and P < 0.05 for replication) for the 3,414 brain imaging phenotypes in allele frequency, gene proximity, and the number of linkage disequilibrium proxies using the reference panels of EUR (1KGP) (a) and EAS (1KGP) (b), respectively. The dashed line in red is the number of trait-related loci colocalized with the 6,361 index variants for brain imaging phenotypes identified in this study. We compute empirical significance ( P < 0.05) by tallying the number of sets showing the same or more colocalized loci than the colocalized loci with index variants of brain imaging phenotypes identified by this study. ExtendedDataFig5.tif Distribution of the CHIMGEN participants (n = 7,306) across centers and MRI scanners. a, The number of participants recruited from each center. b, The percentage of participants whose MRI data are acquired by each type of MRI scanners. The discovery sample is defined as the participants (70.60%, blue) whose MRI data are acquired by GE DISCOVERY MR750 with the same parameters. The replication sample is defined as other participants (29.40%, green) whose MRI data are acquired by other scanners. GE DISCOVERY MR750* represents neuroimaging data collected by the GE DISCOVERY MR750 scanner but using different scanning parameters from the discovery sample. ExtendedDataFig6.tif Filtration strategy of the CHIMGEN participants. After excluding participants without blood sample or failed to pass QC of genomic or neuroimaging data, we finally included 7,058 participants in the sMRI analyses, 6,982 participants in the dMRI analyses, and 6,310 participants in the rs-MRI analyses. dMRI, diffusion MRI; MRI, magnetic resonance imaging; QC, quality control; rs-fMRI, resting-state functional MRI; and sMRI, structural MRI. ExtendedDataFig7.tif Performance of ComBat harmonization. a, b, Correlation matrices before (a) and after (b) ComBat harmonization show inter-scanner correlations of surface areas across the 74 left cerebral cortical subregions derived from the Destrieus (a2009s) parcellation in one representative participant who traveled to and was scanned at 28 MR scanners. c, d, Correlation matrices before (c) and after (d) harmonization show inter-scanner correlations of L1 across the 48 white matter fiber tracts derived from the JHU atlas in the same participant. e, Distributions of surface areas of the 74 subregions in the 30 MR scanners before (upper row) and after (lower row) harmonization. f, Distributions of L1 of the 48 white matter fiber tracts in the 30 MR scanners before (upper row) and after (lower row) harmonization. Notes: MRI data of center 15 are acquired at center 1 and MRI data of center 20 are acquired at center 2. ExtendedDataFig8.tif Quality control of CHIMGEN genetic data. After a series of variant-level and sample-level quality control procedures, 549,309 variants and 7,163 participants are finally included in this study. Hetero, heterozygosity; HWE, Hardy-Weinberg equilibrium; M, mean; MAF, minor allele frequency; PCA, principal component analysis; SD, standard deviation. The participant deviated from the whole population in PCA also has an excessive heterozygosity (greater than M+5SD). ExtendedDataFig9.tif Quality assessments of CHIMGEN genetic data. a, Genotype concordance (> 99.79%) of non-missing calls between 86 pairs of duplicate samples indicates high genotyping reproducibility. b, Genotype heterozygosity and missing rates of 549,309 variants in 7,195 CHIMGEN participants. Missing rates of 4 participants are greater than 3% and heterozygosity rates of 15 participants are greater than five times standard deviations from the mean. c, Correlation of allele frequency across 524,924 overlapping variants between CHIMGEN (n = 7,163) and SG10K (n = 4,810). Bin is colored according to the log10-scaled number of variants within the bin. d, MAF distribution of 704,555 variants in 7,191 CHIMGEN participants. The inset figure shows variant counts (n = 142,828) with MAF < 0.001. e, Genetic population stratification tested by principal component analysis (PCA) in 549,309 variants of the 7,191 CHIMGEN participants, and we detect one outlier (a participant) that deviates from the population. MAF, minor allele frequency. Cite Share Download PDF Status: Published Journal Publication published 29 May, 2024 Read the published version in Nature Genetics → Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-2047527","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":139089808,"identity":"f4d66cc7-bf6b-4b8e-a78e-c5eefd30d2dc","order_by":0,"name":"Chunshui Yu","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA00lEQVRIiWNgGAWjYDACCSBmbGDgYWBgPgAROUC8FrbEBpK0AAGPIXFa+Gf3mD38ueOwjMGNnO+PbrYxyPHdSGD8XIDPkjtnzI15zxzmMbiRu7E5t43BWPJGArP0DDxaDCRyzKQZ2xBaEjfcSGBj5iGgRfInWEvOQ5CWeqK0SPBCtDCCtCQYENIicSOtTJq3LZ1H8swzw9k55yQMZ5552CyNTwv/jORtQIdZ2/MdT37wOafMRh7IOPgZnxY4UDgAsZUBFk2EgTyR6kbBKBgFo2AEAgCXqEy+lZ+xdAAAAABJRU5ErkJggg==","orcid":"https://orcid.org/0000-0001-5648-5199","institution":"Tianjin Medical University General Hospital","correspondingAuthor":true,"submittingAuthor":false,"prefix":"","firstName":"Chunshui","middleName":"","lastName":"Yu","suffix":""},{"id":139089809,"identity":"8120fe63-de7e-4667-9cf3-12d97db8e336","order_by":1,"name":"Jilian Fu","email":"","orcid":"","institution":"Tianjin Medical University General Hospital","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Jilian","middleName":"","lastName":"Fu","suffix":""},{"id":139089810,"identity":"eee804c7-16bd-4a6f-bed9-1e723f019eb3","order_by":2,"name":"Quan Zhang","email":"","orcid":"","institution":"Characteristic Medical Center of Chinese People's Armed Police Force","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Quan","middleName":"","lastName":"Zhang","suffix":""},{"id":139089811,"identity":"6e35fb07-216e-4181-ac2c-6cc4f7383d48","order_by":3,"name":"Jianhua Wang","email":"","orcid":"","institution":"School of Basic Medical Sciences, Tianjin Medical University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Jianhua","middleName":"","lastName":"Wang","suffix":""},{"id":139089812,"identity":"3490ef63-05ae-4e44-a114-7109ebb6ead6","order_by":4,"name":"Meiyun Wang","email":"","orcid":"https://orcid.org/0000-0003-4454-5005","institution":"Henan Provincial People's Hospital \u0026 the People's Hospital of Zhengzhou University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Meiyun","middleName":"","lastName":"Wang","suffix":""},{"id":139089813,"identity":"f24c68d3-80d3-4fbe-8e24-2ddc2f4d8435","order_by":5,"name":"Bing Zhang","email":"","orcid":"","institution":"Nanjing Drum Tower Hospital","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Bing","middleName":"","lastName":"Zhang","suffix":""},{"id":139089814,"identity":"62ee3255-0a8e-42ca-a380-d73754ffc48a","order_by":6,"name":"Wenzhen Zhu","email":"","orcid":"","institution":"Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Wenzhen","middleName":"","lastName":"Zhu","suffix":""},{"id":139089815,"identity":"eec3fb7a-02a5-43ec-b97f-fd0c0ca97699","order_by":7,"name":"Shijun Qiu","email":"","orcid":"","institution":"First Affiliated Hospital of Guangzhou University of Chinese Medicine","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Shijun","middleName":"","lastName":"Qiu","suffix":""},{"id":139089816,"identity":"f2610314-0c7e-4c38-a872-54c2eabc5d3a","order_by":8,"name":"Zuojun Geng","email":"","orcid":"","institution":"The Second Hospital of Hebei Medical University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Zuojun","middleName":"","lastName":"Geng","suffix":""},{"id":139089817,"identity":"5fd5a706-8d61-4334-a4b7-a24fda48b034","order_by":9,"name":"Guangbin Cui","email":"","orcid":"","institution":"Tangdu Hospital, Air Force Medical University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Guangbin","middleName":"","lastName":"Cui","suffix":""},{"id":139089818,"identity":"74bfea16-8cbd-4238-b4ec-051869fe15e5","order_by":10,"name":"Yongqiang Yu","email":"","orcid":"","institution":"The First Affiliated Hospital of Anhui Medical University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Yongqiang","middleName":"","lastName":"Yu","suffix":""},{"id":139089819,"identity":"11d54d74-0300-4c1b-af89-c621aba20c7b","order_by":11,"name":"Weihua Liao","email":"","orcid":"","institution":"Central South University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Weihua","middleName":"","lastName":"Liao","suffix":""},{"id":139089820,"identity":"fb066c57-255f-4277-8fa8-9133804121d7","order_by":12,"name":"Hui Zhang","email":"","orcid":"","institution":"First Hospital of Shanxi Medical University Department of Radiology","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Hui","middleName":"","lastName":"Zhang","suffix":""},{"id":139089821,"identity":"267ce2b0-5322-4937-a890-d1c2751c7bd9","order_by":13,"name":"Bo Gao","email":"","orcid":"","institution":"Yantai Yuhuangding Hospital","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Bo","middleName":"","lastName":"Gao","suffix":""},{"id":139089822,"identity":"6a90bc39-0568-470b-a19b-65032a02879a","order_by":14,"name":"Xiaojun Xu","email":"","orcid":"https://orcid.org/0000-0002-0127-2812","institution":"Second Affiliated Hospital of Zhejiang University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Xiaojun","middleName":"","lastName":"Xu","suffix":""},{"id":139089823,"identity":"77572448-edcc-4d78-8972-8bbc4ab04aa3","order_by":15,"name":"Tong Han","email":"","orcid":"","institution":"Tianjin Huanhu Hospital","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Tong","middleName":"","lastName":"Han","suffix":""},{"id":139089824,"identity":"0347ab24-2ca2-422c-9d9b-c614aceb1fd1","order_by":16,"name":"Zhengwei Yao","email":"","orcid":"","institution":"Huashan Hospital, Fudan University, P.R. China","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Zhengwei","middleName":"","lastName":"Yao","suffix":""},{"id":139089825,"identity":"15eeac4b-b872-4801-a88e-2a477af72ec8","order_by":17,"name":"Wen Qin","email":"","orcid":"https://orcid.org/0000-0002-9121-8296","institution":"Tianjin Medical University General Hospital","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Wen","middleName":"","lastName":"Qin","suffix":""},{"id":139089826,"identity":"bbc60c24-8c19-48cf-9567-7e8645628cae","order_by":18,"name":"Feng Liu","email":"","orcid":"https://orcid.org/0000-0002-3570-4222","institution":"Tianjin Medical University General Hospital","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Feng","middleName":"","lastName":"Liu","suffix":""},{"id":139089827,"identity":"8b5ea86c-1372-438e-923f-3a81dbc15ab2","order_by":19,"name":"Meng Liang","email":"","orcid":"","institution":"Tianjin Medical University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Meng","middleName":"","lastName":"Liang","suffix":""},{"id":139089828,"identity":"4a547cf1-bcfe-4748-be0b-2d1b45b4c2d1","order_by":20,"name":"Sijia Wang","email":"","orcid":"https://orcid.org/0000-0001-8949-303X","institution":"School of Medical Imaging and Tianjin Key Laboratory of Functional Imaging, Tianjin Medical University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Sijia","middleName":"","lastName":"Wang","suffix":""},{"id":139089829,"identity":"9cc8ca6b-a4f2-410d-a7e0-0d2d252fbe62","order_by":21,"name":"Qiang Xu","email":"","orcid":"","institution":"Tianjin Medical University General Hospital","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Qiang","middleName":"","lastName":"Xu","suffix":""},{"id":139089830,"identity":"74f07d37-4291-450d-b737-1e28f252ef93","order_by":22,"name":"Jiayuan Xu","email":"","orcid":"","institution":"Department of Radiology and Tianjin Key Laboratory of Functional Imaging, Tianjin Medical University General Hospital","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Jiayuan","middleName":"","lastName":"Xu","suffix":""},{"id":139089831,"identity":"b532be3b-cd94-43c8-96cb-fd7ed8ccde9e","order_by":23,"name":"Peng Zhang","email":"","orcid":"","institution":"Tianjin Medical University Cancer Institute and Hospital","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Peng","middleName":"","lastName":"Zhang","suffix":""},{"id":139089832,"identity":"d071d322-7e7f-484a-9f76-6ff754892996","order_by":24,"name":"Wei Li","email":"","orcid":"","institution":"Tianjin Medical University Cancer Institute and Hospital","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Wei","middleName":"","lastName":"Li","suffix":""},{"id":139089833,"identity":"7028372f-96fa-4a28-997c-ccc5cd14b430","order_by":25,"name":"Dapeng Shi","email":"","orcid":"","institution":"Henan Provincial People's Hospital \u0026 Zhengzhou University People's Hospital","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Dapeng","middleName":"","lastName":"Shi","suffix":""},{"id":139089834,"identity":"d529eb69-2049-4d10-a110-e7c0d9db0a12","order_by":26,"name":"Caihong Wang","email":"","orcid":"","institution":"The First Affiliated Hospital of Zhengzhou University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Caihong","middleName":"","lastName":"Wang","suffix":""},{"id":139089835,"identity":"bde9ceb4-91a4-4017-955d-2f9d16181944","order_by":27,"name":"Su Lui","email":"","orcid":"https://orcid.org/0000-0003-3541-1769","institution":"West China Hospital of Sichuan University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Su","middleName":"","lastName":"Lui","suffix":""},{"id":139089836,"identity":"f467a02d-3bfd-4aec-860a-a38bf82c08d4","order_by":28,"name":"Zhihan Yan","email":"","orcid":"","institution":"The Second Affiliated Hospital and Yuying Children's Hospital of Wenzhou Medical University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Zhihan","middleName":"","lastName":"Yan","suffix":""},{"id":139089837,"identity":"cc2861fe-621d-4fee-b91b-e8ce9062a73e","order_by":29,"name":"Feng Chen","email":"","orcid":"https://orcid.org/0000-0002-9129-7895","institution":"Hainan General Hospital","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Feng","middleName":"","lastName":"Chen","suffix":""},{"id":139089838,"identity":"1c76e4c9-5d49-414a-acf2-dfb6da04a2d6","order_by":30,"name":"Jing Zhang","email":"","orcid":"","institution":"Lanzhou University Second Hospital","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Jing","middleName":"","lastName":"Zhang","suffix":""},{"id":139089839,"identity":"14d0c85c-2b4d-414a-83ac-efd1503f6a4c","order_by":31,"name":"Jiance Li","email":"","orcid":"","institution":"The First Affiliated Hospital of Wenzhou Medical University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Jiance","middleName":"","lastName":"Li","suffix":""},{"id":139089840,"identity":"14def595-bc44-4af8-bbb4-3ddd0db24152","order_by":32,"name":"Wen Shen","email":"","orcid":"","institution":"Tianjin First Center Hospital","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Wen","middleName":"","lastName":"Shen","suffix":""},{"id":139089841,"identity":"1e8da869-c4d1-492c-8edc-0ebcdcb2a883","order_by":33,"name":"Yanwei Miao","email":"","orcid":"","institution":"First Affiliated Hospital of Dalian Medical University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Yanwei","middleName":"","lastName":"Miao","suffix":""},{"id":139089842,"identity":"8f22a599-662e-4278-9414-61ace892cf88","order_by":34,"name":"Dawei Wang","email":"","orcid":"","institution":"Qilu Hospital of Shandong University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Dawei","middleName":"","lastName":"Wang","suffix":""},{"id":139089843,"identity":"ef939423-fb5c-4b98-96e2-fbdcf91f39c1","order_by":35,"name":"Junfang Xian","email":"","orcid":"","institution":"Beijing Tongren Hospital, Capital Medical University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Junfang","middleName":"","lastName":"Xian","suffix":""},{"id":139089844,"identity":"761eee1f-dbf5-46b5-b13a-5bc586bae610","order_by":36,"name":"Jia-Hong Gao","email":"","orcid":"","institution":"Academy for Advanced Interdisciplinary Studies, Peking University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Jia-Hong","middleName":"","lastName":"Gao","suffix":""},{"id":139089845,"identity":"eb928103-34b8-4186-8321-bb3d391040b6","order_by":37,"name":"Xiaochu Zhang","email":"","orcid":"https://orcid.org/0000-0002-7541-0130","institution":"University of Science and Technology of China","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Xiaochu","middleName":"","lastName":"Zhang","suffix":""},{"id":139089846,"identity":"35f96bcf-ef8f-4363-9620-b8d779ff6759","order_by":38,"name":"Kai Xu","email":"","orcid":"","institution":"The Affiliated Hospital of Xuzhou Medical University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Kai","middleName":"","lastName":"Xu","suffix":""},{"id":139089847,"identity":"3ff403e0-55d2-4e80-8860-a7374c4ad37c","order_by":39,"name":"Xi-Nian Zuo","email":"","orcid":"https://orcid.org/0000-0001-9110-585X","institution":"Beijing Normal University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Xi-Nian","middleName":"","lastName":"Zuo","suffix":""},{"id":139089848,"identity":"71546d21-7be5-4a44-8fa0-59725f953438","order_by":40,"name":"Long Jiang Zhang","email":"","orcid":"https://orcid.org/0000-0002-6664-7224","institution":"Jinling Hospital, Medical School of Nanjing University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Long","middleName":"Jiang","lastName":"Zhang","suffix":""},{"id":139089849,"identity":"e4584704-e8c1-4491-b0b1-98185b94c3bb","order_by":41,"name":"Zhaoxiang Ye","email":"","orcid":"https://orcid.org/0000-0003-3157-8393","institution":"Tianjin Medical University Cancer Institute and Hospital","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Zhaoxiang","middleName":"","lastName":"Ye","suffix":""},{"id":139089850,"identity":"b463db87-6d51-4099-b470-c8a4cce9628a","order_by":42,"name":"Jingliang Chen","email":"","orcid":"","institution":"First Affiliated Hospital of Zhengzhou University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Jingliang","middleName":"","lastName":"Chen","suffix":""},{"id":139089851,"identity":"4887073a-474e-4830-94a2-d676fe89a737","order_by":43,"name":"Mulin Jun Li","email":"","orcid":"","institution":"School of Basic Medical Sciences, Tianjin Medical University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Mulin","middleName":"Jun","lastName":"Li","suffix":""}],"badges":[],"createdAt":"2022-09-09 05:50:58","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-2047527/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-2047527/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1038/s41588-024-01766-y","type":"published","date":"2024-05-29T04:00:00+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":26968783,"identity":"e51e6270-9a6c-4664-8b86-7e2faf137cd9","added_by":"auto","created_at":"2022-09-26 13:40:37","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":1869485,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eGenetic discovery of EAS-GWASsin CHIMGEN.\u003c/strong\u003e \u003cstrong\u003ea\u003c/strong\u003e, Ideogram shows genome-wide significant associations (discovered at \u003cem\u003eP\u003c/em\u003e \u0026lt; 5 × 10−8 and \u003cem\u003eP\u003c/em\u003e \u0026lt; 1.46 × 10−11) found in EAS-GWASs for brain imaging phenotypes in CHIMGEN, including 37 novel associations discovered at \u003cem\u003eP\u003c/em\u003e\u0026lt; 1.46 × 10−11. \u003cstrong\u003eb\u003c/strong\u003e, rs111737551 shows significant associations with many brain diffusion imaging phenotypes, of which MD ofthe left anterior limb of internal capsule has the strongest association. Colors represent subgroups of phenotypes and dashed line means a cutoff at Bonferroni-corrected\u003cem\u003e P\u003c/em\u003e \u0026lt; 1.46 × 10−5. \u003cstrong\u003ec\u003c/strong\u003e,Manhattan plot shows genome-wide associations for MD of the left anterior limb of internal capsule, of which rs111737551 has the strongest association. \u003cstrong\u003ed\u003c/strong\u003e, Spatial mapping of rs111737551 against voxel-wise MD (\u003cem\u003eq\u003c/em\u003e \u0026lt; 0.05, FDR corrected) in white matter in 4,969 participants. \u003cstrong\u003ee\u003c/strong\u003e,\u003cstrong\u003e f\u003c/strong\u003e, Examples of novel associations identified by EAS-GWASs. One novel association (\u003cstrong\u003ee\u003c/strong\u003e) is between rs77768175 and the right caudate volume. The top panel displays that the caudate is the most significant region in spatial mapping of rs77768175 against voxel-wise GMV (\u003cem\u003eq\u003c/em\u003e\u0026lt; 0.05, FDR corrected) in gray matter in 5,025 participants. The middle and bottom panels show regional plots of the 500 kb centered at rs77768175 in CHIMGEN and UKBB, respectively. LD values of other variants with rs77768175 are marked with colors, but these values are not available in UKBB because the minor allele frequency of this variant in EUR was zero. Another novel association (\u003cstrong\u003ef\u003c/strong\u003e) is between rs2274224 and functional activity amplitude of the ventral attention RSN. The top panel shows the spatial map of the ventral attention RSN, and the middle and bottom panels show regional plots in CHIMGEN and UKBB. FDR, false discovery rate; GMV, gray matter volume; INFO, information scores; LD, linkage disequilibrium; MD, mean diffusivity; RSN, resting-state network.\u003c/p\u003e","description":"","filename":"Fig1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-2047527/v1/b29c14ef4bed737ce193a0fa.jpg"},{"id":26969499,"identity":"f660b2e0-110e-4be4-8da7-527949b99e85","added_by":"auto","created_at":"2022-09-26 13:45:37","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":1686182,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eGenetic discovery in trans-ancestral GWAS meta-analyses. a\u003c/strong\u003e,\u003cstrong\u003e \u003c/strong\u003eHistogram shows the distribution of 1,677 significant associations identified in trans-ancestral GWAS meta-analyses (discovered at \u003cem\u003eP\u003c/em\u003e \u0026lt; 1.46 × 10−11 and replicated at \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.05) across chromosomes. Blue denotes 1,218 known associations and red represents 459 novel associations defined as the corresponding loci do not overlap with any loci of the same brain imaging phenotype derived from both EAS-GWASs and EUR-GWASs. \u003cstrong\u003eb\u003c/strong\u003e,\u003cstrong\u003e \u003c/strong\u003eGenomic distribution and significance (-log10(\u003cem\u003ep\u003c/em\u003e-value)) of associations. Red represents 459 novel associations and blue represents 1,218 known associations. \u003cstrong\u003ec\u003c/strong\u003e,\u003cstrong\u003e \u003c/strong\u003eHistogramshows the distribution of 1,677 associations across subgroups (different colors) of phenotypes. The dotted part denotes 1,218 known associations and the blanked part denotes 459 novel associations.\u003cstrong\u003e d\u003c/strong\u003e,\u003cstrong\u003e \u003c/strong\u003ePhenotypic distribution and significance (-log10(\u003cem\u003ep\u003c/em\u003e-value)) of associations. Stars mean 459 novel associations and dots mean 1,218 known associations. \u003cstrong\u003ee\u003c/strong\u003e, \u003cstrong\u003ef\u003c/strong\u003e,\u003cstrong\u003e \u003c/strong\u003eExamples for novel associations identified by trans-ancestral GWAS meta-analyses. One is the association of chromosome 7q22.1 with the left cerebellum VIIIb volume (\u003cstrong\u003ee\u003c/strong\u003e) and another is the association of chromosome 7p22.1 with the brain stem volume (\u003cstrong\u003ef\u003c/strong\u003e). These two associations are only significant in trans-ancestral GWAS meta-analyses, but neither in EAS analyses nor in EUR analyses.\u003c/p\u003e","description":"","filename":"Fig2.jpg","url":"https://assets-eu.researchsquare.com/files/rs-2047527/v1/dabf1ae044d1ebab53acc974.jpg"},{"id":26969600,"identity":"dd805c1e-bb48-43f3-94a0-5fa3f5033016","added_by":"auto","created_at":"2022-09-26 13:50:37","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":1500253,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePooled genome-wide significant associations. a\u003c/strong\u003e, Manhattan plot depicts the number of significant associations (discover at \u003cem\u003eP\u003c/em\u003e\u0026lt; 5 × 10−8 and replicate at \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.05) at each variant identified in any of the single-ancestral and trans-ancestral GWASs. \u003cstrong\u003eb\u003c/strong\u003e, Histogram demonstrates the distribution of the pooled 6,361 significant associations across chromosomes. \u003cstrong\u003ec\u003c/strong\u003e,\u003cstrong\u003e \u003c/strong\u003eLollipop chart shows the distribution of the pooled 6,361 associations across subgroups (different colors) of brain imaging phenotypes.\u003cstrong\u003e d\u003c/strong\u003e,\u003cstrong\u003e \u003c/strong\u003eLollipop chart shows the across-subgroup distribution of significant associations normalized by the number of phenotypes in each subgroup. \u003cstrong\u003ee\u003c/strong\u003e,\u003cstrong\u003e \u003c/strong\u003eSpatial distribution of the pooled significant associations across brain regions derived from the Destrieux (a2009s) parcellation. \u003cstrong\u003ef\u003c/strong\u003e,\u003cstrong\u003e \u003c/strong\u003eSpatial distribution of the pooled associations across brain fiber tracts defined by the AutoPtx. \u003cstrong\u003eg\u003c/strong\u003e,\u003cstrong\u003e \u003c/strong\u003eDistribution of the pooled associations for functional activity amplitude (circle color) and functional connectivity (link) of brain RSNs derived from the 25-component group-ICA. \u003cstrong\u003eh\u003c/strong\u003e,\u003cstrong\u003e \u003c/strong\u003eSignificant functional enrichment for genes derived from variants of the pooled significant associations. Fold enrichment larger than 10 is set to 10.\u003c/p\u003e","description":"","filename":"Fig3.jpg","url":"https://assets-eu.researchsquare.com/files/rs-2047527/v1/0aa5fa547920fd9f8ed0af26.jpg"},{"id":26969829,"identity":"7799ac00-ea48-4091-a426-5618fca23d13","added_by":"auto","created_at":"2022-09-26 13:55:37","extension":"jpg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":1426006,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eAncestry-shared and ancestry-specific genetic discovery. a\u003c/strong\u003e,\u003cstrong\u003e \u003c/strong\u003eManhattan plot of the genome-wide associations of L3 of the left anterior corona radiata shows both ancestry-shared and ancestry-specific associations. \u003cstrong\u003eb\u003c/strong\u003e, Pie (all imaging modalities) and circle (each imaging modality) charts show the numbers and/or ratios of ancestry-shared (discover and replicate at \u003cem\u003eP\u003c/em\u003e ≥ 0.05 in CQ-tests), ancestry-specific (discover at \u003cem\u003eP\u003c/em\u003e\u0026lt; 1.02 × 10−5 and replicate at \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.05) and un-classified associations in the 4,890 pooled associations. \u003cstrong\u003ec\u003c/strong\u003e, Portions of the ancestry-shared and ancestry-specific associations with and without validation when using comparable sample sizes in EAS (n = 7,058 from CHIMGEN) and EUR (n = 8,428 from UKBB). \u003cstrong\u003ed\u003c/strong\u003e, Brain regions with ancestry-specific associations for brain structural imaging phenotypes. \u003cstrong\u003ee\u003c/strong\u003e, Brain white matter tracts with ancestry-specific associations for brain diffusion imaging phenotypes. \u003cstrong\u003ef\u003c/strong\u003e, Functional activity amplitude (circle color) in and functional connectivity (link) between brain RSNs with ancestry-specific associations for brain functional imaging phenotypes. No colored circle indicates all functional activity amplitude of RSNs without ancestry-specific associations. \u003cstrong\u003eg, h, i,\u003c/strong\u003e Examples for ancestry-specific and ancestry-shared associations. \u003cstrong\u003e(g)\u003c/strong\u003e is an EAS-specific association between rs8081528 and FA of the left medial lemniscus; \u003cstrong\u003e(h)\u003c/strong\u003e is an EUR-specific association between rs2923402 and the left pallidum volume; and \u003cstrong\u003e(i)\u003c/strong\u003e is an ancestry-shared association between rs13164785 and MD of the left uncinate fasciculus.\u003c/p\u003e","description":"","filename":"Fig4.jpg","url":"https://assets-eu.researchsquare.com/files/rs-2047527/v1/5bbb22f65c563d8e69fcc148.jpg"},{"id":26969830,"identity":"c9ff9f10-f1b7-45b5-9eb8-7bae844c5f1a","added_by":"auto","created_at":"2022-09-26 13:55:37","extension":"jpg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":1719301,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eGenetic colocalizations of brain imaging phenotypes with brain-related non-imaging traits. a\u003c/strong\u003e, The left panel demonstrates the number of genome-wide significant loci (\u003cem\u003eP\u003c/em\u003e \u0026lt; 5 × 10−8) identified by GWAS for each of the 37 brain-related non-imaging traits. The orange part shows the number of loci colocalized with brain imaging phenotypes, which is defined as a trait-related locus containing at least one index variant of the pooled significant associations for brain imaging phenotypes discovered at \u003cem\u003eP\u003c/em\u003e \u0026lt; 5 × 10−8 and replicated at \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.05. The middle-left panel shows significance (\u003cem\u003eq\u003c/em\u003e\u0026lt; 0.05, FDR corrected) in Fisher’s exact test detecting whether the probability of a trait-related locus colocalized with brain imaging phenotypes is significantly higher than the probability of any evenly segmented genomic region of 500 kb (similar size with trait-related locus) colocalized with brain imaging phenotypes. The other two panels show significance (\u003cem\u003eP\u003c/em\u003e \u0026lt; 0.05) in the 1,000 resampling tests detecting whether the identified number of colocalized trait-related loci (orange color in the left panel) with the index variants of brain imaging phenotypes is significantly higher than the number of colocalized trait-related loci with randomly generated 1,000 sets of variants with matched allele frequency, gene proximity, and the number of LD proxies using the EUR and EAS of 1KGP as reference panels, respectively. \u003cstrong\u003eb\u003c/strong\u003e, Ideogram shows the portion of colocalized loci of the non-imaging traits whose index variants showing a high LD (r2 \u0026gt; 0.8) with the index variants of the brain imaging phenotypes. Since rs13107325(chromosome 4q24) is a colocalized variant between 5 non-imaging traits and 107 brain structural imaging phenotypes, we only depict 30 colocalizations between each trait and structural phenotypes at this locus for readability and clarity. \u003cstrong\u003ec,\u003c/strong\u003e An example of colocalization demonstrates that rs13107325 associated with the left putamen volume is also linked to cognitive performance, intelligence, risky behaviors, drinks per week, and schizophrenia. In the y-axis for the volume of the left putamen, we depict \u003cem\u003ep\u003c/em\u003e-value from trans-ancestral GWAS meta-analysis if variants exist in both CHIMGEN and UKBB, and \u003cem\u003ep\u003c/em\u003e-value from EAS-GWAS or EUR-GWAS if variants exist only in CHIMGEN or UKBB.\u003c/p\u003e","description":"","filename":"Fig5.jpg","url":"https://assets-eu.researchsquare.com/files/rs-2047527/v1/6f7da07ba04647a46566a1fe.jpg"},{"id":57396478,"identity":"ab6224ff-494b-4fc0-9768-c14f50887b53","added_by":"auto","created_at":"2024-05-30 07:06:49","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":10084285,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-2047527/v1/c1916e2b-cfe6-408b-9677-625aa7c9ccc4.pdf"},{"id":26968787,"identity":"ba1425d4-093d-43fd-a9e3-7705012a08a0","added_by":"auto","created_at":"2022-09-26 13:40:37","extension":"xlsx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":5783884,"visible":true,"origin":"","legend":"\u003cp\u003eDataset 1\u003c/p\u003e","description":"","filename":"supplementarytables.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-2047527/v1/c0c9f51aea2d7ec280fe4166.xlsx"},{"id":26968790,"identity":"b7392179-23d3-470d-85a5-2efa72785bee","added_by":"auto","created_at":"2022-09-26 13:40:37","extension":"tif","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":2933016,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eAssociations of rs67827860 with brain diffusion imaging phenotypes in CHIMGEN. \u003c/strong\u003eThe rs67827860 shows significant associations with many brain diffusion imaging phenotypes. Colors of points represent the subgroups of phenotypes. Dashed line means a cutoff at Bonferroni-corrected threshold of \u003cem\u003eP\u003c/em\u003e \u0026lt; 1.46 × 10−5. FA, fractional anisotropy; L1, maximal eigenvalue; L2, medial eigenvalue; L3, minimal eigenvalues; MD, mean diffusivity; MO, mode of anisotropy.\u003c/p\u003e","description":"","filename":"ExtendedDataFig2.tif","url":"https://assets-eu.researchsquare.com/files/rs-2047527/v1/61d0a56d7745bd798aaca3e4.tif"},{"id":26968799,"identity":"a1562ee7-949d-4b22-a0de-dda617447bcd","added_by":"auto","created_at":"2022-09-26 13:40:38","extension":"tif","order_by":4,"title":"","display":"","copyAsset":false,"role":"supplement","size":44929252,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSpatial distribution of pooled independent significant associations in each brain atlas or parcellation. a\u003c/strong\u003e, Diedrichsen cerebellar atlas (SUIT); \u003cstrong\u003eb\u003c/strong\u003e, Harvard-Oxford subcortical atlas; \u003cstrong\u003ec,\u003c/strong\u003eHarvard-Oxford cortical atlas; \u003cstrong\u003ed, \u003c/strong\u003eSubcortical structures from FIRST; \u003cstrong\u003ee, \u003c/strong\u003eSubcortical structures from aseg; \u003cstrong\u003ef, \u003c/strong\u003eBA exvivo parcellation; \u003cstrong\u003eg\u003c/strong\u003e, Desikan-Killiany (DK) parcellation; \u003cstrong\u003eh\u003c/strong\u003e, Desikan-Killiany-Tourville (DKT) parcellation; \u003cstrong\u003ei\u003c/strong\u003e, Parcellation based on the pial surface using Desikan-Killiany parcellation (pial); \u003cstrong\u003ej\u003c/strong\u003e, JHU fibers; \u003cstrong\u003ek\u003c/strong\u003e, UKBB 100-component group-ICA. In \u003cstrong\u003ea-j\u003c/strong\u003e, colors represent the numbers of significant associations for each brain region. In \u003cstrong\u003ek\u003c/strong\u003e, circle colors reflect the numbers of significant associations for functional activity amplitude of each brain RSN and a link indicates the existence of a significant association for functional connectivity between every two brain RSNs.\u003c/p\u003e","description":"","filename":"ExtendedDataFig3.tif","url":"https://assets-eu.researchsquare.com/files/rs-2047527/v1/65d6ef07a153fcb324f6b110.tif"},{"id":26969604,"identity":"7727e2e3-8fb6-40fb-8f80-e386727cf3d6","added_by":"auto","created_at":"2022-09-26 13:50:37","extension":"tif","order_by":5,"title":"","display":"","copyAsset":false,"role":"supplement","size":5800228,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eDistinguishing true genetic sharing between brain-related non-imaging traits and brain imaging phenotypes from colocalizations by chance. a, b, \u003c/strong\u003eFor each brain-related non-imaging trait, a random distribution of colocalizations (blue) is generated by the number of trait-related loci colocalized with each of 1,000 sets of randomly created variants matching with the index variants of 6,361 pooled independent genome-wide significant associations (\u003cem\u003eP\u003c/em\u003e \u0026lt; 5 × 10−8 for discovery and \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.05 for replication) for the 3,414 brain imaging phenotypes in allele frequency, gene proximity, and the number of linkage disequilibrium proxies using the reference panels of EUR (1KGP) (\u003cstrong\u003ea\u003c/strong\u003e) and EAS (1KGP) (\u003cstrong\u003eb\u003c/strong\u003e), respectively. The dashed line in red is the number of trait-related loci colocalized with the 6,361 index variants for brain imaging phenotypes identified in this study. We compute empirical significance (\u003cem\u003eP\u003c/em\u003e \u0026lt; 0.05) by tallying the number of sets showing the same or more colocalized loci than the colocalized loci with index variants of brain imaging phenotypes identified by this study.\u003c/p\u003e","description":"","filename":"ExtendedDataFig4.tif","url":"https://assets-eu.researchsquare.com/files/rs-2047527/v1/d8bf10166bb44f004b5a06ce.tif"},{"id":26969502,"identity":"d83d1d40-b8bc-4283-b847-190955c064bd","added_by":"auto","created_at":"2022-09-26 13:45:37","extension":"tif","order_by":6,"title":"","display":"","copyAsset":false,"role":"supplement","size":250308,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eDistribution of the CHIMGEN participants (n = 7,306) across centers and MRI scanners. a, \u003c/strong\u003eThe number of participants recruited from each center.\u003cstrong\u003e b, \u003c/strong\u003eThe percentage of participants whose MRI data are acquired by each type of MRI scanners.\u003cstrong\u003e \u003c/strong\u003eThe discovery sample is defined as the participants (70.60%, blue) whose MRI data are acquired by GE DISCOVERY MR750 with the same parameters. The replication sample is defined as other participants (29.40%, green) whose MRI data are acquired by other scanners.\u003cstrong\u003e \u003c/strong\u003eGE DISCOVERY MR750* represents neuroimaging data collected by the GE DISCOVERY MR750 scanner but using different scanning parameters from the discovery sample.\u003c/p\u003e","description":"","filename":"ExtendedDataFig5.tif","url":"https://assets-eu.researchsquare.com/files/rs-2047527/v1/95a41f2671cd11f290e606a8.tif"},{"id":26969504,"identity":"a95f7d0c-2a40-4f44-9785-7aa548fdc125","added_by":"auto","created_at":"2022-09-26 13:45:37","extension":"tif","order_by":7,"title":"","display":"","copyAsset":false,"role":"supplement","size":524890,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eFiltration strategy of the CHIMGEN participants.\u003c/strong\u003e After excluding participants without blood sample or failed to pass QC of genomic or neuroimaging data, we finally included 7,058 participants in the sMRI analyses, 6,982 participants in the dMRI analyses, and 6,310 participants in the rs-MRI analyses. dMRI, diffusion MRI; MRI, magnetic resonance imaging; QC, quality control; rs-fMRI, resting-state functional MRI; and sMRI, structural MRI.\u003c/p\u003e","description":"","filename":"ExtendedDataFig6.tif","url":"https://assets-eu.researchsquare.com/files/rs-2047527/v1/00dfc3769725881191923c0c.tif"},{"id":26968791,"identity":"f371b6ee-2b3e-4778-ad51-d08ded13f33d","added_by":"auto","created_at":"2022-09-26 13:40:37","extension":"tif","order_by":8,"title":"","display":"","copyAsset":false,"role":"supplement","size":7188400,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePerformance of ComBat harmonization. a\u003c/strong\u003e, \u003cstrong\u003eb,\u003c/strong\u003e Correlation matrices before (\u003cstrong\u003ea\u003c/strong\u003e) and after (\u003cstrong\u003eb\u003c/strong\u003e) ComBat harmonization show inter-scanner correlations of surface areas across the 74 left cerebral cortical subregions derived from the Destrieus (a2009s) parcellation in one representative participant who traveled to and was scanned at 28 MR scanners. \u003cstrong\u003ec\u003c/strong\u003e, \u003cstrong\u003ed,\u003c/strong\u003e Correlation matrices before (\u003cstrong\u003ec\u003c/strong\u003e) and after (\u003cstrong\u003ed\u003c/strong\u003e) harmonization show inter-scanner correlations of L1 across the 48 white matter fiber tracts derived from the JHU atlas in the same participant. \u003cstrong\u003ee\u003c/strong\u003e, Distributions of surface areas of the 74 subregions in the 30 MR scanners before (upper row) and after (lower row) harmonization. \u003cstrong\u003ef\u003c/strong\u003e, Distributions of L1 of the 48 white matter fiber tracts in the 30 MR scanners before (upper row) and after (lower row) harmonization.\u003c/p\u003e\n\u003cp\u003eNotes: MRI data of center 15 are acquired at center 1 and MRI data of center 20 are acquired at center 2.\u003c/p\u003e","description":"","filename":"ExtendedDataFig7.tif","url":"https://assets-eu.researchsquare.com/files/rs-2047527/v1/abd7a2520ec5b25b2461c8aa.tif"},{"id":26969603,"identity":"bdd5dc9e-0a66-4df2-8ccc-77e9eec03a36","added_by":"auto","created_at":"2022-09-26 13:50:37","extension":"tif","order_by":9,"title":"","display":"","copyAsset":false,"role":"supplement","size":449140,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eQuality control of CHIMGEN genetic data.\u003c/strong\u003e After a series of variant-level and sample-level quality control procedures, 549,309 variants and 7,163 participants are finally included in this study. Hetero, heterozygosity; HWE, Hardy-Weinberg equilibrium; M, mean; MAF, minor allele frequency; PCA, principal component analysis; SD, standard deviation.\u003c/p\u003e\n\u003cp\u003e* The participant deviated from the whole population in PCA also has an excessive heterozygosity (greater than M+5SD).\u003c/p\u003e","description":"","filename":"ExtendedDataFig8.tif","url":"https://assets-eu.researchsquare.com/files/rs-2047527/v1/19a30ef50c8dc6340e248714.tif"},{"id":26968796,"identity":"cffcd967-950f-4231-867f-5804dc617ca7","added_by":"auto","created_at":"2022-09-26 13:40:37","extension":"tif","order_by":10,"title":"","display":"","copyAsset":false,"role":"supplement","size":10545832,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eQuality assessments of CHIMGEN genetic data. a\u003c/strong\u003e, Genotype concordance (\u0026gt; 99.79%) of non-missing calls between 86 pairs of duplicate samples indicates high genotyping reproducibility. \u003cstrong\u003eb\u003c/strong\u003e, Genotype heterozygosity and missing rates of 549,309 variants in 7,195 CHIMGEN participants. Missing rates of 4 participants are greater than 3% and heterozygosity rates of 15 participants are greater than five times standard deviations from the mean. \u003cstrong\u003ec\u003c/strong\u003e, Correlation of allele frequency across 524,924 overlapping variants between CHIMGEN (n = 7,163) and SG10K (n = 4,810). Bin is colored according to the log10-scaled number of variants within the bin. \u003cstrong\u003ed\u003c/strong\u003e, MAF distribution of 704,555 variants in 7,191 CHIMGEN participants. The inset figure shows variant counts (n = 142,828) with MAF \u0026lt; 0.001. \u003cstrong\u003ee\u003c/strong\u003e, Genetic population stratification tested by principal component analysis (PCA) in 549,309 variants of the 7,191 CHIMGEN participants, and we detect one outlier (a participant) that deviates from the population. MAF, minor allele frequency.\u003c/p\u003e","description":"","filename":"ExtendedDataFig9.tif","url":"https://assets-eu.researchsquare.com/files/rs-2047527/v1/d981db857e86ac06bc9014d2.tif"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"Trans-ancestral genome-wide association studies of brain imaging phenotypes","fulltext":[{"header":"Introduction","content":"\u003cp\u003eBrain imaging phenotypes reflect the structure, function and connectivity of the brain, and they are indicative of cognitive performance and vulnerability of neuropsychiatric disorders\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u003c/sup\u003e. Quantitative brain imaging phenotypes show high heritability\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e and thus the investigation of genetic architecture of the human brain will shed light on the causes for individual differences in cognitive processing and mechanisms of neuropsychiatric disorders. In the past decades, many neuroimaging genetics studies have explored the genetic associations of the brain structure and function, of which several large-scale genome-wide association studies (GWASs) in individuals of European ancestry (EUR) provide unbiased insight into the genetic architectures of brain imaging phenotypes\u003csup\u003e\u003cspan additionalcitationids=\"CR4 CR5 CR6\" citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eThe human brain structure, function and connectivity and the prevalence of inherited brain disorders may differ by ancestries. For example, there are significant differences in regional cortical volume, thickness, and surface area between individuals of East Asian ancestry (EAS) and EUR\u003csup\u003e\u003cspan additionalcitationids=\"CR9\" citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e. In central nervous system demyelinating disorders, Caucasians show higher incidence of multiple sclerosis\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e but lower incidence of neuromyelitis optica spectrum disorder\u003csup\u003e\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e than Asians. Although non-genetic factors may contribute to the differences in brain properties and disorders among ancestries\u003csup\u003e\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u003c/sup\u003e, genetic architectures such as effect size, linkage disequilibrium (LD), and allele frequency (AF) contribute a lot to these differences among ancestries\u003csup\u003e\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003ePrevious GWASs have successfully identified thousands of genetic associations with brain imaging phenotypes\u003csup\u003e\u003cspan additionalcitationids=\"CR4 CR5 CR6\" citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e, yet non-EUR populations are severely under-represented, which prevents us to differentiate ancestry-specific from ancestry-shared genetic associations with brain imaging phenotypes. The knowledge may improve our understanding of ancestry-shared and ancestry-specific relationships between genetic variations and brain disorders since the brain structure and function are intermediate phenotypes linking genetic variations to neuropsychiatric disorders. Despite trans-ancestral GWASs could discover novel variant-phenotype associations and distinguish ancestry-specific from ancestry-shared associations\u003csup\u003e\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e,\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u003c/sup\u003e, trans-ancestral GWASs for brain imaging phenotypes are still lacking.\u003c/p\u003e \u003cp\u003eThe main obstacle to trans-ancestral GWASs of brain imaging phenotypes is the lack of large-scale non-EUR cohorts with both genomic and neuroimaging data, but this dilemma will be broken by the emergence of the Chinese Imaging Genetics (CHIMGEN) cohort\u003csup\u003e\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u003c/sup\u003e (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://chimgen.tmu.edu.cn\u003c/span\u003e\u003cspan address=\"http://chimgen.tmu.edu.cn\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). This cohort has collected genomic and neuroimaging data from 7,306 healthy Chinese Han participants, from which we generated 6,830,145 genomic variants and 3,414 brain imaging phenotypes that have been included in GWASs of EUR individuals from the UK Biobank (UKBB) dataset\u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e. We performed EAS-GWASs for these brain imaging phenotypes based on CHIMGEN data and trans-ancestral GWASs based on GWAS summary statistics of the 3,414 brain imaging phenotypes from EAS (CHIMGEN) and EUR (UKBB) populations. Here, we were interested in: (a) detecting novel genetic associations by incorporating EAS population; (b) exploring the distribution and function of all loci associated with brain imaging phenotypes; (c) identifying ancestry-shared and ancestry-specific associations; and (d) investigating genetic sharing between brain imaging phenotypes and other brain-related traits such as cognition, personality, and neuropsychiatric disorders. A schematic summary is shown in Extended Data Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eGenetic discovery in EAS-GWASs\u003c/h2\u003e \u003cp\u003eIn 7,058 EAS (Chinese Han) participants from the CHIMGEN study, we conducted the first non-EUR GWASs for 3,414 brain imaging phenotypes at the 6,830,145 autosomal variants with a minor allele frequency (MAF)\u0026thinsp;\u0026gt;\u0026thinsp;1%, in which genetic effects were estimated with respect to the number of copies of the non-reference allele. In the discovery dataset of 5,025 EAS participants, we identified 647 genome-wide significant associations (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;5 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;8\u003c/sup\u003e) between genetic variants and brain imaging phenotypes (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ea, Supplementary Table\u0026nbsp;1), and 295 associations were confirmed at \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05 with the same direction of effect in the replication dataset of 2,033 EAS participants. Using a threshold of \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;1.46 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;11\u003c/sup\u003e to additionally correct for the 3,414 GWASs, we found 133 significant variant-phenotype associations in the discovery dataset (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ea, Supplementary Table\u0026nbsp;2). Of the 133 associations, 124 were replicated at \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05 with the same direction of effect in the replication dataset, 124 were also confirmed with a false discovery rate (FDR) of less than 0.05, and 63 survived the Bonferroni-adjusted significance threshold of \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;3.75 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;4\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eThese significant associations were unevenly distributed across the genome, of which the chromosomes 5, 11 and 12 showed more associations with brain imaging phenotypes (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ea). Consistent with the reported pleiotropic variant of rs67827860 at 5q14.3 in EUR participants\u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e, this variant also showed extensive associations with brain diffusion imaging phenotypes in EAS participants (Extended Data Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). In addition, we found another pleiotropic variant of rs111737551 at 11p11.2 that also showed significant associations with various brain diffusion imaging phenotypes (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eb-d). This indel variant is located at an intron of \u003cem\u003eCD82\u003c/em\u003e, which regulates oligodendrocyte progenitor migration and white matter myelination\u003csup\u003e\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u003c/sup\u003e, and mediates age-related cognitive decline\u003csup\u003e\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eTo test the potential of EAS-GWASs in identifying new associations that have not been reported in EUR-GWASs for the same brain imaging phenotypes at the same statistical thresholds, we defined a reference list of the known associations as those with \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;1.46 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;11\u003c/sup\u003e in the discovery dataset (22,138 EUR participants) and \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05 in the replication dataset (11,086 EUR participants) in the currently largest EUR-GWASs for the 3,414 brain imaging phenotypes\u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e, resulting in 1,478 associations (Supplementary Table\u0026nbsp;3). Here, we defined a novel association for EAS-GWASs if the locus of the index variant did not overlap with any loci of the same brain imaging phenotype in the reference list of the known associations. Of the 124 associations (discovered at \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;1.46 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;11\u003c/sup\u003e and replicated at \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05) in EAS-GWASs, 37 (29.8%) associations were considered as novel by this definition (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ea, Supplementary Table\u0026nbsp;4), indicating that GWASs conducted in the non-EUR populations (such as Chinese Han) can reveal novel variant-phenotype associations that are absent in EUR-GWASs even with a larger sample size. For example, rs77768175 at 12q24.13 demonstrated significant association with the right caudate volume (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ee) only in EAS-GWAS. This polymorphic variant in EAS [MAF\u0026thinsp;=\u0026thinsp;0.1607 in 1000 Genomes Project phase 3 (1KGP)] shows far less polymorphic in EUR (MAF\u0026thinsp;=\u0026thinsp;0 in 1KGP), which may explain the absence of signal in EUR-GWAS. Besides, this is an intron variant of \u003cem\u003eHECTD4\u003c/em\u003e which has been associated with epilepsy in EAS but not in EUR\u003csup\u003e\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u003c/sup\u003e. Another EAS-specific association was the correlation of rs2274224 at 10q23.33 with functional activity amplitude of the ventral attention resting-state network (RSN) obtained from the independent component analysis (ICA) (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ef). This missense variant affects PLCE1 protein coding and has been associated with migraine\u003csup\u003e\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003eGenetic discovery in trans-ancestral GWAS meta-analyses\u003c/h2\u003e \u003cp\u003eAt 5,950,889 autosomal variants included in both CHIMGEN and UKBB datasets, trans-ancestral GWAS meta-analyses were conducted based on the fixed-effect model for the 3,414 brain imaging phenotypes shared by the two datasets. In the discovery stage, trans-ancestral GWASs were performed based on summary statistics from the discovery stage of the EAS-GWASs (5,025 CHIMGEN participants) and EUR-GWASs (22,138 UKBB participants). In the replication stage, trans-ancestral analyses were conducted based on summary statistics from the replication stage of the EAS-GWASs (2,033 CHIMGEN participants) and EUR-GWASs (11,086 UKBB participants).\u003c/p\u003e \u003cp\u003eIn the discovery dataset (5,025 EAS and 22,138 EUR), trans-ancestral GWAS meta-analyses revealed 6,920 genome-wide significant variant-phenotype associations (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;5 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;8\u003c/sup\u003e) (Supplementary Table\u0026nbsp;5), of which 5,065 associations were confirmed at \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05 in the replication dataset (2,033 EAS and 11,086 EUR). Using a Bonferroni-adjusted threshold of \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;1.46 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;11\u003c/sup\u003e, we still found 1,746 significant variant-phenotype associations in the discovery dataset (Supplementary Table\u0026nbsp;6). Of these 1,746 associations, we confirmed 1,677 associations at \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05 in the replication dataset, 1,675 at an FDR-adjusted threshold of \u003cem\u003eq\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05, and 1,076 at a Bonferroni-adjusted threshold of \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;2.86 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;5\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eAmong the 1,677 variant-phenotype associations identified in the two-stage trans-ancestral GWAS meta-analyses (discovered at \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;1.46 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;11\u003c/sup\u003e and replicated at \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05), 459 (27.4%) were considered as novel because the locus of the index variant did not overlap with any loci for the same brain imaging phenotype derived from either EAS-GWASs or EUR-GWASs (Supplementary Table\u0026nbsp;7). These novel associations were unevenly distributed across the genome and 32 subgroups of brain imaging phenotypes, of which chromosomes 3, 5 and 11 (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ea and \u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eb) and subgroups of cortical volume, surface area and white matter diffusion (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ec and \u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ed) showed more novel associations. An example is the novel association at 7q22.1 with the left cerebellum VIIIb volume (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ee). The index variant rs1627052 is located at an intron of \u003cem\u003eRELN\u003c/em\u003e, which regulates neuronal migration and cortical layering in the brain, and mutations of this gene are associated with autosomal recessive lissencephaly with cerebellar hypoplasia\u003csup\u003e\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e\u003c/sup\u003e. Another example is the novel association at 7p22.1 with the brain stem volume (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ef). The index variant rs2640 is a missense mutation of \u003cem\u003eEIF2AK1\u003c/em\u003e, showing relations with developmental delay, leukoencephalopathy, and neurologic decompensation\u003csup\u003e\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003ePooled genome-wide significant associations\u003c/h2\u003e \u003cp\u003eWe pooled all genome-wide significant associations (discovered at \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;5 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;8\u003c/sup\u003e and replicated at \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05) for the 3,414 brain imaging phenotypes identified by any of the EAS-GWASs, EUR-GWASs, or trans-ancestral GWASs. We totally identified 6,361 independent associations after merging the overlapping genetic loci derived from the same brain imaging phenotype (Supplementary Table\u0026nbsp;8). Of these 6,361 associations, 1,952 were still significant at \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;1.46 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;11\u003c/sup\u003e with an additional correction for the 3,414 GWASs in the discovery dataset. These 6,361 associations were unevenly distributed across the genome with chromosomes 5 and 17 showing more associations (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ea and \u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eb). These associations spanned all subgroups of brain imaging phenotypes with the subgroups of cortical volume, surface area and white matter diffusion showing more associations (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ec), even considering the different numbers of phenotypes in different subgroups (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ed). For each subgroup of brain imaging phenotypes derived from the same brain atlas or parcellation, significant associations also showed uneven spatial distribution across the brain (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ee-g and Extended Data Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). For example, we observed more associations in the calcarine sulcus, cuneus, precuneus and lingual gyrus for structural imaging phenotypes of the cerebral cortex (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ee), in the forceps minor, superior and inferior longitudinal fasciculi and anterior thalamic radiation for diffusion imaging phenotypes (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ef), in the salience RSN for functional activity amplitude (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eg), and in the basal ganglion RSN for functional connectivity (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eg).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eGenetic variants of the 6,361 independent significant associations were assigned to 856 protein-coding genes (Supplementary Table\u0026nbsp;9) based on the criteria of the location of a variant within 10 kb around a gene. Using all protein-coding genes (n\u0026thinsp;=\u0026thinsp;20,589) as the background, we identified 39 functional enrichment terms in Reactome Pathway Database with an FDR-corrected \u003cem\u003eq\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05 (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eh and Supplementary Table\u0026nbsp;10). The term with the most significant enrichment was nervous system development (FDR-corrected \u003cem\u003eq\u003c/em\u003e\u0026thinsp;=\u0026thinsp;4.74 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;6\u003c/sup\u003e), and the relevant pathways included axon guidance (FDR-corrected \u003cem\u003eq\u003c/em\u003e\u0026thinsp;=\u0026thinsp;4.81 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;6\u003c/sup\u003e), the regulation of commissural axon pathfinding by SLIT and ROBO (FDR-corrected \u003cem\u003eq\u003c/em\u003e\u0026thinsp;=\u0026thinsp;5.29 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;3\u003c/sup\u003e) and others. Another category of terms with significant enrichment was signal transduction (FDR-corrected \u003cem\u003eq\u003c/em\u003e\u0026thinsp;=\u0026thinsp;7.68 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;6\u003c/sup\u003e) involved in neuronal differentiation and migration, such as CRMPs in Sema3A signaling for axonal outgrowth (FDR-corrected \u003cem\u003eq\u003c/em\u003e\u0026thinsp;=\u0026thinsp;2.34 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;2\u003c/sup\u003e), Netrin-1 signaling for axon guidance (FDR-corrected \u003cem\u003eq\u003c/em\u003e\u0026thinsp;=\u0026thinsp;5.28 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;5\u003c/sup\u003e), and NCAM signaling for neurite outgrowth (FDR-corrected \u003cem\u003eq\u003c/em\u003e\u0026thinsp;=\u0026thinsp;1.50 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;2\u003c/sup\u003e). These pooled genome-wide significant loci provide a hitherto largest catalog of genetic associations for human brain imaging phenotypes in diverse populations.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003eAncestry-specific and ancestry-shared genetic discovery\u003c/h2\u003e \u003cp\u003eFrom the GWAS results of the same brain imaging phenotypes conducted in EAS and EUR\u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e, both ancestry-shared and ancestry-specific associations were observed. Taking minimal diffusivity (L3) of the left anterior corona radiata as an example, this brain imaging phenotype was associated with 5q14.3 in both EUR and EAS (ancestry-shared), but with 11p11.2 only in EAS (EAS-specific) and with 17q21.31 only in EUR (EUR-specific) (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ea). To systematically investigate the ancestry-shared and ancestry-specific genetic associations with brain imaging phenotypes, Cochran\u0026rsquo;s Q-test (CQ-test) was used to quantify the heterogeneity of effect size of the pooled 4,890 independent genome-wide significant associations (discovered at \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;5 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;8\u003c/sup\u003e and replicated at \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05) whose index variants were included in both EAS-GWASs and EUR-GWASs. The CQ-test for discovery was conducted based on GWAS summary statistics from the discovery samples of EAS (n\u0026thinsp;=\u0026thinsp;5,025) and EUR (n\u0026thinsp;=\u0026thinsp;22,138) and the CQ-test for replication was performed based on GWAS summary statistics from the replication samples of EAS (n\u0026thinsp;=\u0026thinsp;2,033) and EUR (n\u0026thinsp;=\u0026thinsp;11,086). Ancestry-shared associations were defined as those of \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026ge;\u0026thinsp;0.05 in CQ-tests for both discovery and replication and ancestry-specific associations were defined as those discovered at a Bonferroni-adjusted threshold of \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;1.02 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;5\u003c/sup\u003e and replicated at \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFrom 4,890 variant-phenotype associations, we identified 3,797 homogeneous (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026ge;\u0026thinsp;0.05) and 72 heterogeneous (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;1.02 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;5\u003c/sup\u003e) associations between EAS and EUR in the discovery samples. Of these 3,797 homogeneous associations, 3,524 were replicated at \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026ge;\u0026thinsp;0.05 and considered as ancestry-shared associations (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eb and Supplementary Table\u0026nbsp;11). Of these 72 heterogeneous associations, 43 were replicated at \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05 and considered as ancestry-specific associations (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eb and Supplementary Table\u0026nbsp;11); 38 associations were still significant at an FDR-adjusted threshold of \u003cem\u003eq\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05, and 10 at a Bonferroni-adjusted threshold of \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;6.94 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;4\u003c/sup\u003e. The brain imaging phenotypes from all three imaging modalities consistently showed more ancestry-shared associations than ancestry-specific associations, and the portion (1.76%) of ancestry-specific associations for functional imaging phenotypes was five times greater than that (0.31%) for diffusion imaging phenotypes (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eb).\u003c/p\u003e \u003cp\u003eTo account for the bias due to unbalanced sample sizes between EAS and EUR, we also validated the identified ancestry-shared and ancestry-specific associations in EAS (n\u0026thinsp;=\u0026thinsp;7,058 from CHIMGEN) and EUR (n\u0026thinsp;=\u0026thinsp;8,428 from UKBB)\u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e with comparable sample sizes. After excluding variant-phenotype associations absent in the EUR-GWASs\u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e, we conducted CQ-tests for the remaining 1,845 ancestry-shared associations and 27 ancestry-specific associations. We observed 1,750 (94.85%) ancestry-shared associations at \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026ge;\u0026thinsp;0.05 and 27 ancestry-specific associations (100%) at \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05 (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ec). Of the 27 ancestry-specific associations, all were also replicated at an FDR-corrected threshold of \u003cem\u003eq\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05 and 25 at a Bonferroni-corrected threshold of \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;1.85 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;3\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eIn the structural imaging phenotypes, ancestry-specific associations were mainly observed in the bilateral putamen and precentral gyri, left pallidum and cerebellum VIIIb, and right fusiform gyrus and superior parietal cortex (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ed). In the diffusion imaging phenotypes, ancestry-specific associations were presented in fiber tracts such as the bilateral superior cerebellar peduncles, genu of corpus callosum, right superior corona radiata, and left medial lemniscus (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ee). In the functional imaging phenotypes, ancestry-specific associations were found in four functional connectivity involving seven brain RSNs derived from the ICA (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ef).\u003c/p\u003e \u003cp\u003eThree typical examples of ancestry-shared and ancestry-specific associations are presented in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eg-i. Briefly, an EAS-specific association was uncovered between rs8081528 (17q12) and fractional anisotropy (FA) of the left medial lemniscus (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eg). This variant is located at an intron of \u003cem\u003eCDK12\u003c/em\u003e, which regulates axonal elongation, neurogenesis, and neuronal migration\u003csup\u003e\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e\u003c/sup\u003e. An EUR-specific association was observed between rs2923402 (8p11.21) and left pallidum volume (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eh). The rs2923402 is an intergenic variant close to \u003cem\u003eCHRNB3\u003c/em\u003e, which encodes subunits of nicotinic acetylcholine receptors and is associated with nicotine dependence\u003csup\u003e\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u003c/sup\u003e. An ancestry-shared association was identified between rs13164785 (5q14.3) and mean diffusivity (MD) of the left uncinate fasciculus (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ei). This variant is located at an intron of \u003cem\u003eVCAN\u003c/em\u003e, which encodes a large chondroitin sulfate proteoglycan, a major component of the extracellular matrix. The protein VCAN is involved in cell adhesion, proliferation, migration, and angiogenesis, and is critical for tissue morphogenesis and maintenance\u003csup\u003e\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003eColocalizations with brain-related non-imaging traits\u003c/h2\u003e \u003cp\u003eFor the 6,361 pooled independent genome-wide significant associations (discovered at \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;5 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;8\u003c/sup\u003e and replicated at \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05) of the 3,414 brain imaging phenotypes, we tested if brain imaging phenotypes had shared genetic architectures with non-imaging traits associated with the human brain. Of the 37 brain-related non-imaging traits with available large-scale GWAS summary statistics (Supplementary Table\u0026nbsp;12), 30 traits had at least one genome-wide significant locus containing an index variant of the pooled significant associations for brain imaging phenotypes. To test whether a trait-related locus has greater probability of colocalizing with brain imaging phenotypes than a genomic region of similar size (500 kb), which was generated by evenly segmenting the genome into chunks of 500 kb. The Fisher\u0026rsquo;s exact test (FDR-corrected \u003cem\u003eq\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05) revealed that the associated loci of 19 traits had greater probability of colocalizing with brain imaging phenotypes than any evenly segmented genomic region. Then we used the resampling strategy to assess whether the number of colocalizations of trait-related loci with the index variants of brain imaging phenotypes is significantly higher than the number of colocalizations of these loci with the randomly generated 6,361 variants (1,000 sets). Each set of the created 6,361 variants was matched with the index variants of brain imaging phenotypes in allele frequency, gene proximity, and the number of LD proxies using the EUR and EAS from 1KGP as reference panels, respectively. We found that 16 traits showed larger number of colocalizations than random (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05) in both ancestries and an additional trait was significant only in EAS. The 16 traits with shared genetic architectures with brain imaging phenotypes included cognitive performance, educational attainment, intelligence, neuroticism, risky behaviors, risk tolerance, insomnia symptom, drinks per week, smoking initiation, bipolar disorder, major depressive disorder, schizophrenia, any stroke, migraine, multiple sclerosis, and Parkinson\u0026rsquo;s disease (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003ea and Extended Data Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e). To determine the specific loci shared by each pair of brain imaging phenotype and brain-related non-imaging trait, we only included the loci of the non-imaging trait whose index variants showed a high LD (r\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;\u0026gt;\u0026thinsp;0.8) with the index variants of the brain imaging phenotype and found 1,866 colocalizations based on this criterion (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eb and Supplementary Table\u0026nbsp;13). For example, rs13107325 associated with the left putamen volume was also linked to five non-imaging traits including cognitive performance, intelligence, risky behaviors, drinks per week, and schizophrenia (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003ec). This missense mutation encodes SLC39A8 protein, which mediates the cellular uptake of zinc and manganese, two divalent metal cations that are important for development, tissue homeostasis and immunity\u003csup\u003e\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"Conclusions","content":"\u003cp\u003eWith the neuroimaging genetics data of 7,058 healthy Chinese Han participants from the CHIMGEN study and 33,224 EUR individuals from the UKBB project\u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e, we conducted EAS-GWASs and trans-ancestral GWAS meta-analyses for 3,414 brain imaging phenotypes. The results deepen our understanding of the genetic architectures of brain imaging phenotypes by providing 496 novel associations that were discovered at \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;1.46 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;11\u003c/sup\u003e and replicated at \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05 and highlight the importance of a more global representation of populations in neuroimaging genetics studies. The summary statistics of these GWASs based on Chinese Han participants are freely available on the CHIMGEN website (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://chimgen.tmu.edu.cn/pheweb/\u003c/span\u003e\u003cspan address=\"http://chimgen.tmu.edu.cn/pheweb/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e), which provides public access to browse associations by variant, gene, or phenotype. This website was built with PheWeb (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://github.com/statgen/pheweb/\u003c/span\u003e\u003cspan address=\"https://github.com/statgen/pheweb/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eBy pooling genome-wide significant associations for brain imaging phenotypes in either single-ancestral or trans-ancestral GWASs, we identified 6,361 independent significant associations unevenly distributed in the genome and across subgroups of neuroimaging phenotypes and brain regions, which delineates a landscape of genetic effects on brain structure and function. These genetic associations with brain imaging phenotypes provide candidates for exploring causal genetic mechanisms for brain structure and function and identifying causal genome-brain-disorder pathways\u003csup\u003e\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e\u003c/sup\u003e. The functional enrichment results may improve our understanding of how genomic variations regulate the structural and functional properties of the human brain.\u003c/p\u003e \u003cp\u003eA unique contribution of this study to the field of neuroimaging genetics is the differentiation of ancestry-specific associations from ancestry-shared associations. Consistent with prior studies on non-imaging traits\u003csup\u003e\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e,\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u003c/sup\u003e, brain imaging phenotypes also showed more ancestry-shared associations than ancestry-specific associations, indicating that most of the genetic associations for brain imaging phenotypes in one ancestral population can be extrapolated to other populations. However, we also provide reliable evidence for 43 ancestry-specific associations, which may account for the inter-ancestral difference in brain imaging phenotypes\u003csup\u003e\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e. These ancestry-specific associations are also potential targets for studying inter-ancestral differences of other brain-related traits including brain disorders.\u003c/p\u003e \u003cp\u003eAnother valuable aspect of this study is the finding of colocalizations between brain imaging phenotypes and many brain-related non-imaging traits including cognition, personality, behavior, addiction, and neuropsychiatric disorders, which may improve our understanding of genetic mechanisms underlying the associations between brain imaging and non-imaging phenotypes. These results are useful for the development of diagnostic and therapeutic approaches of neuropsychiatric disorders\u003csup\u003e\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e\u003c/sup\u003e, especially valuable for mental disorders in which their diagnoses are based on a descriptive collection of behaviors without any objective test to stratify patients\u003csup\u003e\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e\u003c/sup\u003e. Genetic loci shared by brain imaging phenotypes and brain disorders are more worthy of biological validation and mechanistic studies, which can accelerate the discovery of novel biomarkers for diagnosis and new targets for treatment. Leveraging genetic instruments, Mendelian randomization can identify brain imaging phenotypes that are causally associated with mental disorders\u003csup\u003e\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e\u003c/sup\u003e, which are more reliable neuroimaging markers of mental disorders than those derived from the intergroup comparisons of neuroimaging data between patients and controls.\u003c/p\u003e"},{"header":"Methods","content":"\u003ch2\u003eParticipants\u003c/h2\u003e \u003cp\u003e\u003cb\u003eEAS participants.\u003c/b\u003e All participants of East Asian ancestry (EAS) were recruited from the CHIMGEN study (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://chimgen.tmu.edu.cn/\u003c/span\u003e\u003cspan address=\"http://chimgen.tmu.edu.cn/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) that has collected genomic, environmental, neuroimaging, and behavioral data from 7,306 healthy Chinese Han participants aged 18\u0026ndash;30 years from 32 research centers located in 21 mainland cities of China from 2015 to 2019. The information and distribution of participants across centers is presented in Supplementary Table\u0026nbsp;14 and Extended Data Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003ea. The CHIMGEN study was reviewed and approved by the Medical Research Ethics Committee of Tianjin Medical University General Hospital and was further reviewed and approved by corresponding local ethics committee of each research center, and written informed consent was obtained from each participant. The inclusion and exclusion criteria of the CHIMGEN participants are provided in Supplementary Table\u0026nbsp;15. The following criteria were further applied to filter the EAS participants by excluding: (a) participants without blood sample for genotyping (n\u0026thinsp;=\u0026thinsp;111); (b) participants who failed to pass the quality control (QC) of genomic data (n\u0026thinsp;=\u0026thinsp;32); and (c) participants who failed to pass the QC of magnetic resonance imaging (MRI) data [n\u0026thinsp;=\u0026thinsp;105 for structural MRI (sMRI); n\u0026thinsp;=\u0026thinsp;181 for diffusion MRI (dMRI); and n\u0026thinsp;=\u0026thinsp;853 for resting-state functional MRI (rs-fMRI)] (Extended Data Fig.\u0026nbsp;6). The EAS participants scanned by the same type of MRI scanners (GE Discovery MR750) with the same parameters were defined as the discovery dataset to reduce the bias resulted from inconsistency in MRI data acquisition, and the remaining participants scanned by other scanners or parameters were defined as the replication dataset to replicate the discovered findings. In genome-wide association studies (GWASs), we finally included 7,058 EAS participants (5,025 for discovery and 2,033 for replication) for brain structural imaging phenotypes, 6,982 EAS participants (4,969 for discovery and 2,013 for replication) for brain diffusion imaging phenotypes, and 6,310 EAS participants (4,645 for discovery and 1,665 for replication) for brain functional imaging phenotypes. The demographic characteristics of the finally included EAS participants for GWASs for brain imaging phenotypes calculated from each MRI modality are listed in Supplementary Table\u0026nbsp;16, and the number of the included EAS participants from each research center for GWASs of brain imaging phenotypes obtained from each MRI modality is presented in Supplementary Table\u0026nbsp;14.\u003c/p\u003e \u003cp\u003e\u003cb\u003eEUR participants.\u003c/b\u003e All participants of European ancestry (EUR) were recruited from the UK Biobank (UKBB) study\u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e,\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e,\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e\u003c/sup\u003e since this study has collected similar genomic and neuroimaging (sMRI, dMRI, and rs-fMRI) data as the CHIMGEN study in more than 30,000 EUR participants. GWAS summary statistics of brain imaging phenotypes from two prior studies with different sample sizes\u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e,\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e were used in this study with different purposes. The dataset with more brain imaging phenotypes (3,414 can be obtained from the CHIMGEN study) and more participants (33,224 EUR participants: 22,138 for discovery and 11,086 for replication)\u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e was used in the main analyses to enhance statistical power. The dataset with 2,640 of the 3,414 brain imaging phenotypes obtained from 8,428 EUR participants\u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e was used to reduce the potential bias resulting from the difference in sample size between EAS and EUR populations.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eImaging data processing in CHIMGEN\u003c/h2\u003e \u003cp\u003e \u003cb\u003eImaging data acquisition.\u003c/b\u003e In the CHIMGEN study, brain imaging data were acquired with ten types of 3.0-Tesla MRI scanners and twelve sets of scanning parameters. The sMRI data were used to calculate brain imaging phenotypes that characterize the structural properties of the cerebrum, cerebellum, brain stem, and subcortical structures; the dMRI data were used to calculate brain imaging phenotypes that reflect diffusion properties of brain white matter tracts; and the rs-fMRI data were used to calculate brain imaging phenotypes that represent brain functional activity amplitude in and functional connectivity between resting-state networks (RSNs). Detailed scanning parameters for each MRI modality for different types of scanners are provided in Supplementary Tables\u0026nbsp;17\u0026ndash;19, and the distribution of participants across types of scanners is presented in Extended Data Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eb.\u003c/p\u003e \u003cp\u003e \u003cb\u003eImaging data QC before preprocessing.\u003c/b\u003e To obtain high-quality brain MRI data, a series of QC procedures were applied before and after the acquisition of brain MRI data in the CHIMGEN study. For example, we optimized scanning parameters for each MRI scanner before acquisition and identified and excluded MRI data with visible lesions, anatomical abnormalities, imaging artefacts, parameter inconsistency, and incomplete brain coverage immediately after acquisition.\u003c/p\u003e \u003cp\u003e \u003cb\u003eImaging data preprocessing.\u003c/b\u003e To extract brain imaging phenotypes accurately and reliably, we developed a series of standardized pipelines to preprocess the multi-modal MRI data from the CHIMGEN study. These pipelines integrated the state-of-the-art preprocessing procedures and tools with the multi-cluster parallel computation. The total computing time was greatly reduced by using the Tianhe super-computer (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.nscc-tj.cn\u003c/span\u003e\u003cspan address=\"https://www.nscc-tj.cn\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). The specific preprocessing pipeline for each type of the MRI data were as follows:\u003c/p\u003e \u003cp\u003e \u003cspan type=\"BoldItalic\" class=\"BoldItalic\" name=\"Emphasis\"\u003esMRI data preprocessing for voxel-based morphometry (VBM).\u003c/span\u003e The CAT12 software (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://dbm.neuro.uni-jena.de/cat\u003c/span\u003e\u003cspan address=\"http://dbm.neuro.uni-jena.de/cat\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) was used to preprocess sMRI data for the VBM analyses. The VBM preprocessing steps included:\u003c/p\u003e \u003cp\u003e(1) Bias correction: Image inhomogeneity caused by B1-field bias was corrected to segment the brain tissue accurately.\u003c/p\u003e \u003cp\u003e(2) Segmentation: The bias-corrected structural MR images were segmented into gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF) using a model based on an adaptive Maximum A Posterior (MAP) technique\u003csup\u003e\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e\u003c/sup\u003e, which does not need a priori information about tissue probabilities.\u003c/p\u003e \u003cp\u003e(3) Creating population-specific tissue templates: To improve the performance of image spatial normalization, the population-specific tissue probability templates for GM, WM, and CSF in the Montreal Neurological Institute (MNI) space were derived from 6,000 CHIMGEN participants using the Diffeomorphic Anatomical Registration Through Exponentiated Lie Algebra (DARTEL) algorithm\u003csup\u003e\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e\u003c/sup\u003e, which was implemented in Statistical Parametric Mapping 12 (SPM12) (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.fil.ion.ucl.ac.uk/spm/software/spm12/\u003c/span\u003e\u003cspan address=\"https://www.fil.ion.ucl.ac.uk/spm/software/spm12/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e(4) Spatial normalization: The segmented GM images were spatially normalized to the population-specific GM template using the DARTEL algorithm and were resampled into a cubic voxel of 1.5 mm. Modulation was then performed on the normalized GM images to preserve the absolute GM volume (GMV).\u003c/p\u003e \u003cp\u003e(5) Phenotype extraction (n\u0026thinsp;=\u0026thinsp;143): From each participant, we extracted the total volumes of GM, WM, CSF, and GM\u0026thinsp;+\u0026thinsp;WM based on brain tissue segmentation and GMVs of 96 cerebral cortical subregions, 28 cerebellar subregions, and 15 subcortical nuclei and brain stem (Supplementary Table\u0026nbsp;20).\u003c/p\u003e \u003cp\u003e \u003cspan type=\"BoldItalic\" class=\"BoldItalic\" name=\"Emphasis\"\u003esMRI data preprocessing for surface-based morphometry (SBM).\u003c/span\u003e FreeSurfer v6.0.0 (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://surfer.nmr.mgh.harvard.edu/\u003c/span\u003e\u003cspan address=\"http://surfer.nmr.mgh.harvard.edu/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) was used to preprocess sMRI data for the SBM analyses with default settings. Specifically, the SBM preprocessing steps included:\u003c/p\u003e \u003cp\u003e(1) Skull stripping: An automated skull-stripping was performed to separate the brain from non-brain tissues in structural MR images. Intensity normalization was applied before and after skull stripping to correct for the intensity non-uniformity due to variations in the sensitivity of reception coils and gradient-driven eddy currents.\u003c/p\u003e \u003cp\u003e(2) Tissue segmentation: A series of tissue segmentation procedures were conducted based on intensity and neighbor constraints to generate the subcortical structures and the boundary between GM and WM.\u003c/p\u003e \u003cp\u003e(3) Surface reconstruction: A two-dimensional tessellated mesh was constructed based on the WM-GM boundary to generate the WM surface, and the WM surface was extended outwards by tracking the GM intensity gradient to generate the pial surface. Topology correction was performed to repair topological defects.\u003c/p\u003e \u003cp\u003e(4) Metric calculation: Surface-based metrics including the cortical thickness, surface area, and cortical volume were calculated based on the pial and WM surfaces.\u003c/p\u003e \u003cp\u003e(5) Estimating spherical normalization parameters: Individual surfaces were then inflated into a spherical space and normalized to the fsaverage template to obtain the spherical normalization parameters.\u003c/p\u003e \u003cp\u003e(6) Phenotype extraction (n\u0026thinsp;=\u0026thinsp;1,035): Surface-based metrics for cortical regions defined by the surface atlases were extracted after converting them from standard to individual space using the inverse spherical normalization parameters. From each participant, we extracted the thickness of 306 cortical regions, the surface area of 370 cortical regions, the volume of 304 cortical regions and 55 subcortical regions (Supplementary Table\u0026nbsp;20).\u003c/p\u003e \u003cp\u003e \u003cspan type=\"BoldItalic\" class=\"BoldItalic\" name=\"Emphasis\"\u003esMRI data preprocessing for subcortical segmentation.\u003c/span\u003e FMRIB\u0026rsquo;s Integrated Registration and Segmentation Tool (FIRST) (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FIRST\u003c/span\u003e\u003cspan address=\"https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FIRST\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) was used to segment subcortical structures with default settings. The preprocessing steps included:\u003c/p\u003e \u003cp\u003e(1) Intensity normalization: We corrected for the intensity non-uniformity due to variations in the sensitivity of reception coils and gradient-driven eddy currents.\u003c/p\u003e \u003cp\u003e(2) Estimating spatial normalization parameters: A two-stage affine registration was performed to convert the intensity-normalized images of each participant to the MNI152 space to obtain spatial normalization parameters.\u003c/p\u003e \u003cp\u003e(3) Segmentation: The inverse normalization parameters were used to convert shape models embedded in FIRST (provided by the Center for Morphometric Analysis) from standard to individual space where the segmentation was performed. Based on these converted models, FIRST searched for the most probable surface mesh of each subcortical structure according to the intensities from the input intensity-normalized images. Then mesh-based subcortical structures were converted to boundary corrected volumetric subcortical structures.\u003c/p\u003e \u003cp\u003e(4) Phenotype extraction (n\u0026thinsp;=\u0026thinsp;15): Based on the FIRST segmentation, we extracted the volumes of brain stem and 14 subcortical regions from each participant.\u003c/p\u003e \u003cp\u003e \u003cspan type=\"BoldItalic\" class=\"BoldItalic\" name=\"Emphasis\"\u003edMRI data preprocessing.\u003c/span\u003e dMRI data were preprocessed using FMRIB Software Library (FSL, version 5.0.10; \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://www.fmrib.ox.ac.uk/fsl\u003c/span\u003e\u003cspan address=\"http://www.fmrib.ox.ac.uk/fsl\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) with the following steps:\u003c/p\u003e \u003cp\u003e(1) Skull stripping: The non-brain tissues were removed from the b\u0026thinsp;=\u0026thinsp;0 images to generate a binary mask for the following tensor metric calculation.\u003c/p\u003e \u003cp\u003e(2) Motion and distortion correction: The \u003cem\u003eeddy_openmp\u003c/em\u003e program was used to evaluate and repair image displacement and signal dropout caused by head motion, and image distortion caused by eddy current.\u003c/p\u003e \u003cp\u003e(3) Tensor metric calculation: The linear least square algorithm was used to estimate the diffusion tensor and to calculate diffusion metrics of each voxel from the tensor using the DTIFIT program. The obtained diffusion metrics included the three eigenvalues (L1, L2, and L3), mean diffusivity (MD), fractional anisotropy (FA), and mode of anisotropy (MO).\u003c/p\u003e \u003cp\u003e(4) Estimating spatial normalization parameters: A two-step procedure was used to estimate the normalization parameters between individual diffusion and MNI standard space. Specifically, individual b\u0026thinsp;=\u0026thinsp;0 images were aligned to structural images using the Boundary-Based Registration (BBR) algorithm. The obtained BBR transformation matrix was then concatenated with the DARTEL deformation field from individual to MNI space generated in the VBM preprocessing. The merged deformation field and its inverse deformation field were finally used in the following analyses.\u003c/p\u003e \u003cp\u003e(5) Probabilistic fiber tracking: The BEDPOSTX program was used to estimate the diffusion orientation distribution based on a ball-stick model with the following parameters: maximum number of fibers per voxel\u0026thinsp;=\u0026thinsp;3, burn-in period\u0026thinsp;=\u0026thinsp;1,000, number of iterations\u0026thinsp;=\u0026thinsp;1,250, and deconvolution model\u0026thinsp;=\u0026thinsp;sticks with a range of diffusivities. After converting the pre-defined seed, target, exclusion, and stop masks of AutoPtx (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://fsl.fmrib.ox.ac.uk/fsl/fslwiki/AutoPtx\u003c/span\u003e\u003cspan address=\"https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/AutoPtx\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) from standard to individual diffusion space using the inverse deformation field, the probabilistic fiber tracking was performed using the PROBTRACKX program with these converted masks in the individual diffusion space with the parameters of number of samples\u0026thinsp;=\u0026thinsp;5,000 and angle threshold\u0026thinsp;=\u0026thinsp;80 degree.\u003c/p\u003e \u003cp\u003e(6) Metric normalization: The diffusion metrics were normalized into the MNI space using the above-mentioned merged deformation field and resampled into a cubic voxel of 2 mm.\u003c/p\u003e \u003cp\u003e(7) Estimating diffusion metrics on white matter skeleton: We used a modified tract-based spatial statistics (TBSS) pipeline to create the white matter skeleton. In contrast to the standard TBSS pipeline\u003csup\u003e\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e\u003c/sup\u003e that directly aligns the individual FA images to the averaged FA template (FMRIB-58) in MNI space using the FNIRT program, we normalized individual FA images using the merged deformation field. Then a mean FA image of all individuals was created and \u0026ldquo;thinned\u0026rdquo; to generate a mean white matter skeleton representing the centers of white matter tracts common to all individuals. The aligned FA images of each participant were then projected onto the mean white matter skeleton by filling the mean skeleton with FA values from the nearest tract center, which was achieved by searching perpendicular to the local skeleton structure for maximal value. The obtained projection parameters were applied to other diffusion metric images (L1, L2, L3, MD, and MO) of the participant to estimate the diffusion properties of each voxel on the white matter skeleton.\u003c/p\u003e \u003cp\u003e(8) Phenotype extraction (n\u0026thinsp;=\u0026thinsp;450): From each participant, we extracted the L1, L2, L3, MD, FA, and MO of 75 white matter tracts (Supplementary Table\u0026nbsp;20).\u003c/p\u003e \u003cp\u003e \u003cspan type=\"BoldItalic\" class=\"BoldItalic\" name=\"Emphasis\"\u003ers-fMRI data preprocessing\u003c/span\u003e. The rs-fMRI data preprocessing included the following steps:\u003c/p\u003e \u003cp\u003e(1) Discarding unstable volumes: Since rs-fMRI data were acquired using three repetition times (TRs: 0.71, 0.8, and 2 seconds) and the functional images acquired at each TR were defined as a functional volume, the first functional volumes (\u0026asymp;\u0026thinsp;10 seconds: 15, 13, 5 volumes for different TRs) were discarded to allow signal to reach equilibrium and to ensure the participants to adapt to scanning noise.\u003c/p\u003e \u003cp\u003e(2) Slice timing correction: The remaining volumes were corrected for intra-volume temporal differences using sinc-interpolation.\u003c/p\u003e \u003cp\u003e(3) Head motion correction: Inter-volume head motion was corrected by realigning each volume to the mean volume using a six-parameter rigid-body transformation. The frame-wise displacement (FD) was calculated by the Jenkinson approach\u003csup\u003e\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e\u003c/sup\u003e to index volume-to-volume changes in head position. When the FD of a volume was greater than 0.5 mm, this volume and its one previous volume and two subsequent volumes were defined as affected volumes.\u003c/p\u003e \u003cp\u003e(4) Spatial normalization: After removing non-brain tissues from the head-motion-corrected functional images, the obtained functional images were co-registered to the structural images using the BBR method. Then all co-registered functional volumes were spatially normalized to the MNI space using the deformation field obtained from the VBM preprocessing and resampled to 3-mm isotropic voxels.\u003c/p\u003e \u003cp\u003e(5) Noise reduction and bandpass filtering: Nuisance covariates including linear trend, Friston-24 head motion parameters, affected volumes, and WM and CSF signals were regressed out, and temporal bandpass filtering (0.01\u0026ndash;0.08 Hz) was applied to reduce low-frequency drift and high-frequency noise.\u003c/p\u003e \u003cp\u003e(6) Defining RSNs: RSNs were defined as the kept components from two group independent component analyses (group-ICA) conducted based on UKBB data\u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e. We also kept 21 RSNs from the 25-component ICA and 55 RSNs from the 100-component ICA.\u003c/p\u003e \u003cp\u003e (7) Back-reconstruct: A dual-regression method was used to back-reconstruct RSNs of each participant.\u003c/p\u003e \u003cp\u003e(8) Defining functional metrices: After regressing out characteristic time courses of the non-RSN components, the characteristic time courses of RSNs were used to define two functional metrics. The functional activity amplitude of each RSN was defined as the standard deviation of fluctuations of the noise-removed characteristic time course of the RSN. The functional connectivity was defined as the temporal correlation of the noise-removed characteristic time courses between every two RSNs.\u003c/p\u003e \u003cp\u003e(9) Phenotype extraction (n\u0026thinsp;=\u0026thinsp;1,771): From each participant, we extracted 76 phenotypes to represent functional activity amplitude of each RSN and 1,695 phenotypes to reflect functional connectivity between every two RSNs (Supplementary Table\u0026nbsp;20).\u003c/p\u003e \u003cp\u003e \u003cspan type=\"BoldItalic\" class=\"BoldItalic\" name=\"Emphasis\"\u003eQC of the preprocessed MRI data.\u003c/span\u003e We also checked the preprocessed MRI data to find errors or imperfections emerged during imaging data preprocessing. If they were identified in a participant, we first tried to find reasons for imperfection and to re-run the pipeline after fixing them. If the newly obtained preprocessed MR images were still problematic, we had to exclude the participant from the following analyses for brain imaging phenotypes derived from the problematic imaging modality. The errors and imperfections mainly included bad tissue segmentation, imperfect spatial normalization, incorrect non-brain tissue removal, intensity normalization error, pial surface misplacement, topological defect, and fiber tracking error. The SPM12 was used to evaluate head motion in the rs-fMRI data. If the maximum displacement in any of the three orthogonal directions was more than 3.0 mm or a maximum rotation was greater than 3.0 degree, the rs-fMRI data of this participant would be excluded. We also excluded rs-fMRI data of participants with mean FD\u0026thinsp;\u0026gt;\u0026thinsp;0.5 mm or with affected volumes more than one third of the total volumes.\u003c/p\u003e \u003cp\u003e \u003cb\u003eExtraction of brain imaging phenotypes.\u003c/b\u003e To perform trans-ancestral GWAS meta-analyses of brain imaging phenotypes, we generated 3,414 brain imaging phenotypes from EAS participants (CHIMGEN), all of which were also included in the previous GWASs in EUR participants (UKBB)\u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e. These included 1,193 phenotypes of volume, surface area, and cortical thickness generated from sMRI data, 1,771 phenotypes of functional activity amplitude and functional connectivity obtained from rs-fMRI data, and 450 phenotypes of brain diffusion properties derived from dMRI data. A list of the included 3,414 brain imaging phenotypes is provided in Supplementary Table\u0026nbsp;20.\u003c/p\u003e \u003cp\u003e \u003cb\u003eHarmonization and normalization of brain imaging phenotypes.\u003c/b\u003e The use of different scanners and parameters may bring bias to the integrated analyses of brain imaging phenotypes obtained from multiple centers. Therefore, we applied ComBat harmonization to the 3,414 brain imaging phenotypes to remove variations of phenotypes resulting from scanner and parameter differences while preserving biologically relevant information\u003csup\u003e\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e\u003c/sup\u003e. Here, we first assessed the performance of ComBat harmonization in two subjects who traveled to and were scanned at 28 centers. For each brain imaging measure, the inter-scanner consistency was assessed by the correlations of this measure across brain regions between every two scanners in each subject. Brain imaging measures acquired from different MR scanners showed different degrees of inter-scanner inconsistencies before harmonization; however, the inter-scanner consistencies of these measures were greatly improved after ComBat harmonization (Extended Data Fig.\u0026nbsp;7a-d). Then, we calculated the distributions of each brain imaging phenotype across participants for each MR scanner before and after ComBat harmonization. We found that the distributions of brain imaging phenotypes became more similar across different MR scanners after harmonization (Extended Data Fig.\u0026nbsp;7e and 7f).\u003c/p\u003e \u003cp\u003eThe distributions of brain imaging phenotypes varied considerably even after harmonization, with a portion of phenotypes showing skewed distribution that would violate the assumption of the normal distribution of the phenotypic data when using the linear regression model to perform GWAS analyses. Thus, normal score transformation was applied to brain imaging phenotypes to make the data normally distributed and reduce undue influence of outliers. Specifically, the real values of each brain imaging phenotype were ranked from lowest to highest and these ranks were matched to and then replaced by equivalent ranks of random numbers generated from a normal distribution.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003eImaging data processing in UKBB\u003c/h2\u003e \u003cp\u003eThe acquisition, preprocessing, and quality control of brain MRI data, and the extraction, harmonization, and normalization of brain imaging phenotypes have been described elsewhere\u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e,\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e\u003c/sup\u003e. The data preprocessing and metric calculation approaches for the 3,414 brain imaging phenotypes were similar between CHIMGEN and UKBB.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003eGenetic data processing in CHIMGEN\u003c/h2\u003e \u003cp\u003e \u003cb\u003eBlood collection, DNA extraction and genotyping.\u003c/b\u003e After blood sample collection, centrifugation and isolation were applied immediately in each research center to obtain plasma and buffy coat of each participant, which were then transported to the Tianjin Medical University General Hospital by a professional biomedical cold chain logistics company for centralized management and unified preprocessing. After standardized storage, the buffy coat was delivered to a sequencing company (Novogene Bioinformatics Technology, \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://en.novogene.com/\u003c/span\u003e\u003cspan address=\"https://en.novogene.com/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) for DNA extraction and genotyping. The CWE2100 Blood DNA Kit was used to extract DNA following the manufacturers\u0026rsquo; specifications. In participants with qualified DNA samples defined as greater than 10 ng/\u0026micro;l of concentration and 260/280 between 1.8 and 2.2 of purity, the Asian Screening Array 750K (ASA-750K) specially designed for Asian populations was applied to 1ug DNA to capture genome-wide genetic variations. With one plate position for one sample, a 96-position plate can simultaneously deal with 96 DNA samples. To validate the reproducibility of genotyping, we deliberately included 86 blind duplicates (one duplicate for one plate) in the experiment. We then calculated the concordance rate of the genotyping results for every duplicate pair and found high concordance rates (ranging from 99.79\u0026ndash;99.97%) (Extended Data Fig.\u0026nbsp;9a).\u003c/p\u003e \u003cp\u003e \u003cb\u003ePre-imputation QC for genotyped data.\u003c/b\u003e Given the high homogeneity in racial identity (Chinese Han) in the CHIMGEN participants, a putative QC pipeline was used for the genotyped data with PLINKv2.0\u003csup\u003e39\u003c/sup\u003e (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://zzz.bwh.harvard.edu/plink/\u003c/span\u003e\u003cspan address=\"http://zzz.bwh.harvard.edu/plink/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). The QC procedures for CHIMGEN genetic data (7,195 participants genotyped at 743,722 variants) are presented in Extended Data Fig.\u0026nbsp;8, and the specific procedures were as follows:\u003c/p\u003e \u003cp\u003e(1) Sex checking: We used genotyped data from the X chromosome to infer the sex of each participant and compared the result with the sex reported by the participant. We defined the inconsistency between the inferred and reported sex as sex mismatch, which is possibly caused by sample mishandling or DNA contamination. Since males only have one copy of X chromosome, they are expected to be homozygous for X markers outside the pseudo-autosomal region. Thus, the homozygosity rate of the X chromosome should be greater than 0.8 in males and smaller than 0.2 in females\u003csup\u003e\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e\u003c/sup\u003e. Based on these criteria, we found and excluded 2 participants with sex mismatch.\u003c/p\u003e \u003cp\u003e(2) Sample-level missing rate: After excluding 17,555 duplicated variants from the 743,722 genotyped variants, we computed the genotype missing rate for each participant using the \u003cem\u003e--miss\u003c/em\u003e command in PLINK in the remaining 726,167 variants. We found that 4 out of the 7,195 participants had a missing rate greater than 3% (Extended Data Fig.\u0026nbsp;9b), and these participants were excluded from the following analyses.\u003c/p\u003e \u003cp\u003e(3) Variant-level missing rate: We also calculated genotype call rate for each of the 726,167 variants in the 7,191 participants. We found 21,612 variants with a missing rate\u0026thinsp;\u0026gt;\u0026thinsp;5% (a genotype call rate\u0026thinsp;\u0026lt;\u0026thinsp;95%), and these variants were excluded.\u003c/p\u003e \u003cp\u003e(4) Minor allele frequency (MAF): We also computed MAF for each of the remaining 704,555 variants in the 7,191 participants. We found and excluded 142,828 variants with MAF\u0026thinsp;\u0026lt;\u0026thinsp;0.001 and retained 561,727 variants for the next QC step (Extended Data Fig.\u0026nbsp;9d).\u003c/p\u003e \u003cp\u003e(5) Hardy-Weinberg equilibrium (HWE): We estimated the deviation of each of the 561,727 variants from HWE and excluded 12,418 variants that were significantly deviated from HWE (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;10\u003csup\u003e\u0026minus;\u0026thinsp;6\u003c/sup\u003e). The retained 549,309 variants were entered into the following QC procedures.\u003c/p\u003e \u003cp\u003e(6) Heterozygosity rate: The extremely high heterozygosity rate is an indicator of poor DNA quality. Heterozygosity is defined by (N\u0026thinsp;\u0026minus;\u0026thinsp;O)/N, where N is the number of non-missing genotypes and O is the observed number of homozygous genotypes for a given participant. Based on the 549,309 variants, we calculated the heterozygosity rate for each of the remaining 7,191 participants using the \u003cem\u003e--het\u003c/em\u003e command in PLINK. We excluded 15 participants with a heterozygosity greater than five times standard deviations from the mean (Extended Data Fig.\u0026nbsp;9b).\u003c/p\u003e \u003cp\u003e(7) Related participants: A requirement of the population-based GWAS is that all included participants are unrelated. If duplicates and relatives are present, a bias will be introduced in the study because the genotypes within families are overrepresented\u003csup\u003e\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e\u003c/sup\u003e. Although we have paid attention to the requirement during the recruitment of participants, the property of multi-center and large-scale dataset made it unavoidable to include unintended duplicates or relatives. To identify duplicate and related individuals, we calculated identity by descent (IBD) for each pair of individuals based on independent variants. After removing the genomic regions with extended linkage disequilibrium (LD) entirely, the remaining genomic regions were pruned until no pair of variants within a given window (100 variants) was correlated (r\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;\u0026gt;\u0026thinsp;0.2). The remaining variants were defined as independent variants. Any pair of participants with an IBD\u0026thinsp;\u0026gt;\u0026thinsp;0.1875 was considered as duplicate or related individuals. We found 11 duplicate or related pairs and then we excluded the one with higher sample-level missing rate from the following analyses.\u003c/p\u003e \u003cp\u003e(8) Population structure: We used principal component analysis (PCA) to capture population structure of the CHIMGEN cohort, and the PCA resulted in 20 principal components (PCs). The PCA was conducted to identify individuals deviated from the Chinese Han population by projecting all participants onto the first two components of the four HapMap3 populations (CEU, CHB, JPT, YRI). We found one participant with extreme deviation from the CHIMGEN cohort (Extended Data Fig.\u0026nbsp;9e), and then this participant was excluded from the following analyses.\u003c/p\u003e \u003cp\u003eIn summary, the variant-level quality control excluded 17,555 duplicated variants, 21,612 variants with missing rate\u0026thinsp;\u0026gt;\u0026thinsp;5%, 12,418 variants significantly deviated from HWE (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;10\u003csup\u003e\u0026minus;\u0026thinsp;6\u003c/sup\u003e), and 142,828 variants with MAF\u0026thinsp;\u0026lt;\u0026thinsp;0.001. Finally, 549,309 variants out of the 743,722 genotyped variants were included in genomic imputation. In the sample-level quality control, we excluded 2 sex mismatching participants, 11 duplicate or related participants, 4 participants with a genotype missing rate greater than 3%, and 15 participants with a heterozygosity greater than five times standard deviations from the mean (1 participant also with outlying population structure). Finally, 7,163 out of the 7,195 participants were included in the following analyses.\u003c/p\u003e \u003cp\u003eTo evaluate genotyping quality of the CHIMGEN participants, we calculated the across-variant correlation of allele frequencies in 524,924 overlapping variants between CHIMGEN and SG10K (another cohort of Asian population)\u003csup\u003e\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e\u003c/sup\u003e. We found a significant correlation (Spearman correlation: r\u0026thinsp;=\u0026thinsp;0.96, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;1.00 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;322\u003c/sup\u003e) in allele frequencies between the two Asian cohorts (Extended Data Fig.\u0026nbsp;9c). This finding indicates high genotyping quality of the CHIMGEN participants, although the CHIMGEN and SG10K cohorts had different sample sizes (7,163 versus 4,810) and slightly different ancestral backgrounds (Chinese Han versus Asian populations) and used different genomic technologies (genotyping versus sequencing).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003eImputation and post-imputation QC in CHIMGEN\u003c/h2\u003e \u003cp\u003eThe haplotypes were estimated based on the genotypes of the 549,309 qualified variants using SHAPEIT2\u003csup\u003e42\u003c/sup\u003e (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://mathgen.stats.ox.ac.uk/genetics_software/shapeit/shapeit.html\u003c/span\u003e\u003cspan address=\"https://mathgen.stats.ox.ac.uk/genetics_software/shapeit/shapeit.html\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) with default parameters. The phased autosomal variants were then imputed using the IMPUTE2\u003csup\u003e43\u003c/sup\u003e (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://mathgen.stats.ox.ac.uk/impute/impute_v2.html\u003c/span\u003e\u003cspan address=\"http://mathgen.stats.ox.ac.uk/impute/impute_v2.html\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) in chunks of 5,000 kb with a combined reference panel that merged 1000 Genomes Project (1KGP, phase 3; n\u0026thinsp;=\u0026thinsp;2,504) and SG10K (n\u0026thinsp;=\u0026thinsp;4,471). We evaluated the imputation performance of the proposed scheme from the following two aspects:\u003c/p\u003e \u003cp\u003e(1) Imputation accuracy among reference panels\u003c/p\u003e \u003cp\u003eWe compared the accuracy of imputation results obtained from three different reference panels. The first reference panel was the 1KGP panel that was downloaded from the website of IMPUTE2 (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://mathgen.stats.ox.ac.uk/impute/1000GP_Phase3.html\u003c/span\u003e\u003cspan address=\"https://mathgen.stats.ox.ac.uk/impute/1000GP_Phase3.html\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). The second reference panel was derived from the phased 4,471 unrelated participants selected from the SG10K data of 4,810 participants. In the 4,810 SG10K participants, the unrelated participants were identified by calculating kinship coefficients (\u0026lt;\u0026thinsp;1/2\u003csup\u003e(9/2)\u003c/sup\u003e) using the KING software\u003csup\u003e\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e\u003c/sup\u003e based on common (MAF\u0026thinsp;\u0026gt;\u0026thinsp;0.05) and relatively independent variants (pruned using PLINK with parameters: \u003cem\u003e--indep-pairwise 1000 80 0.1\u003c/em\u003e). To identify the maximum number of unrelated individuals, we listed all related pairs and iteratively removed individual that appeared most frequently in the list until the list was empty. The third reference panel (1KGP\u0026thinsp;+\u0026thinsp;SG10K) was generated using the following three steps: (a) imputing 1KGP to SG10K; (b) imputing SG10K to 1KGP; and (c) merging the two imputed datasets to construct the 1KGP\u0026thinsp;+\u0026thinsp;SG10K panel using the IMPUTE2 program (\u003cem\u003e-merge_ref_panels\u003c/em\u003e).\u003c/p\u003e \u003cp\u003e To compare imputation accuracy among reference panels, we extracted 43,726 variants on chromosome 2 from 7,163 CHIMGEN participants. The genotype calls of 4,373 variants (1 out of every 10 variants sorted by position) were masked and saved for the evaluation of imputation accuracy. For each reference panel, we pre-phased the chromosome 2 using a reference-based phasing followed by imputation. Imputation error rate was estimated by comparing the imputed genotypes for the masked variants with their genotyped results. In addition, we counted imputed variants with information scores (INFO)\u0026thinsp;\u0026ge;\u0026thinsp;0.8 under four continuous MAF bins (0.005\u0026ndash;0.01, 0.01\u0026ndash;0.05, 0.05\u0026ndash;0.2, and 0.2\u0026ndash;0.5) for the results derived from each reference panel. We found that the 1KGP\u0026thinsp;+\u0026thinsp;SG10K panel had the best imputation performance over the other two panels (1KGP and SG10K). Specifically, the 1KGP\u0026thinsp;+\u0026thinsp;SG10K panel could impute the most high-quality variants (INFO\u0026thinsp;\u0026ge;\u0026thinsp;0.8) across all MAF bins (Extended Data Fig.\u0026nbsp;10a), while keeping the lowest imputation error rate (Extended Data Fig.\u0026nbsp;10b).\u003c/p\u003e \u003cp\u003e(2) Imputation accuracy among genotyping arrays\u003c/p\u003e \u003cp\u003eWe also tested whether the ASA-750K specially designed for Asian populations could improve imputation performance in Chinese participants compared to other three arrays designed for populations with other ancestries, including the Illumina Global Screening Array, Affymetrix EUR, and Affymetrix Biobank that had comparable numbers of variants with the ASA-750K. Based on the genotype calls on chromosome 20 obtained from the high-coverage whole genome sequencing (WGS) data of 90 Chinese participants\u003csup\u003e\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e\u003c/sup\u003e, we extracted the genotypes of variants included in each array to mimic the genotyped data generated by that array. Then the extracted variants were used to impute other variants not included in the array with the 1KGP\u0026thinsp;+\u0026thinsp;SG10K panel. Since these 90 participants had been included in the 1KGP panel, we excluded these participants in the generation of the merged panel. We then compared the consistency of the imputed genotypes based on variants included in different arrays with the WGS genotypes on chromosome 20. The ASA-750K showed the best imputation accuracy for the CHIMGEN genetic data among the four arrays (Extended Data Fig.\u0026nbsp;10c).\u003c/p\u003e \u003cp\u003eFinally, the 549,309 variants were imputed to 111,370,847 autosomal variants with the 1KGP\u0026thinsp;+\u0026thinsp;SG10K panel. Extended Data Fig.\u0026nbsp;10d shows the distribution of INFO on all markers in the imputed dataset. To avoid false positive signals in the association analyses, we filtered variants with MAF\u0026thinsp;\u0026lt;\u0026thinsp;0.01 and INFO\u0026thinsp;\u0026lt;\u0026thinsp;0.9, and finally kept 6,830,145 autosomal variants in the GWASs. The distributions of MAF and INFO of the finally included 6,830,145 autosomal variants are presented in Extended Data Fig.\u0026nbsp;10e.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003eGenetic data processing and imputation in UKBB\u003c/h2\u003e \u003cp\u003eThe detailed processing and imputation steps for the genetic data of UKBB were provided in a previous study\u003csup\u003e\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003eCovariates for EAS-GWASs in CHIMGEN\u003c/h2\u003e \u003cp\u003eSince a variety of confounding factors might mask or bias the effects of genetic variants on brain imaging phenotypes, we controlled for a series of covariates in EAS-GWASs to reduce the risk of reporting false positive associations. We designed covariates according to the previous GWASs for brain imaging phenotypes\u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e,\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u003c/sup\u003e, including age, sex, age \u0026times; sex, age \u0026times; age, and top three PCs for EAS-GWASs for all brain imaging phenotypes. Only the top three PCs were selected as covariates because they could account for the main variance of population stratification in this highly homogeneous population (CHIMGEN) (Extended Data Fig.\u0026nbsp;11). The total intracranial volume (TIV) was also included in GWASs for regional brain structural imaging phenotypes (such as cortical volume, surface area, and cortical thickness), and mean FD (measuring head motion) was included in GWASs for brain functional imaging phenotypes (functional activity amplitude and functional connectivity). All included covariates were transformed into Z-scores and missing values were set to zero in the transformed data. The covariates used in EUR-GWASs in UKBB can be found in previous studies\u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e,\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003eEAS-GWASs for brain imaging phenotypes in CHIMGEN\u003c/h2\u003e \u003cp\u003eFor the yielded 6,830,145 autosomal variants from CHIMGEN, an additive model was applied to investigate the association between the dosage of each variant and each brain imaging phenotype using BGENEv1.2\u003csup\u003e4\u003c/sup\u003e (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://jmarchini.org/bgenie/\u003c/span\u003e\u003cspan address=\"https://jmarchini.org/bgenie/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e), which was specially designed to deal with GWASs of many phenotypes performed simultaneously. GWASs were performed for 3,414 brain imaging phenotypes based on up to 7,058 EAS participants, including 1,193 brain structural imaging phenotypes (5,025 participants for discovery and 2,033 participants for replication), 1,771 brain functional imaging phenotypes (4,645 participants for discovery and 1,665 participants for replication), and 450 brain diffusion imaging phenotypes (4,969 participants for discovery and 2,013 participants for replication), while controlling for the above-mentioned covariates. In the GWAS for each brain imaging phenotype, the independent variant-phenotype associations were identified by the following steps: (a) all variants with \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;5 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;8\u003c/sup\u003e in the discovery sample were used to create a list of variants; (b) a locus of 500 kb centered at the most significant variant (index variant) in the list was generated and all variants within the locus were removed from the list; (c) the remaining variants formed a new list of variants and then the step (b) was repeated; (d) the iterative process stopped until the list was empty; and (e) the iterative process would generate several loci with independent index variants, of which the overlapping loci were merged into an independent locus indexed by the most significant variant of these loci. In this way, we identified all independent variant-phenotype associations in the discovery sample of each GWAS. We also repeated the analyses with a threshold of \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;1.46 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;11\u003c/sup\u003e in the discovery samples to additionally correct for the 3,414 brain imaging phenotypes. In the replication samples, we reported the associations that were confirmed at uncorrected \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05, false positive rate corrected (FDR-corrected) \u003cem\u003eq\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05, and Bonferroni-corrected \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003eTrans-ancestral GWAS meta-analyses for brain imaging phenotypes\u003c/h2\u003e \u003cp\u003eTrans-ancestral GWAS meta-analyses can boost the power to detect novel genetic associations when the underlying causal variants are shared between ancestries. For the 5,950,889 autosomal variants included in the CHIMGEN and UKBB datasets, a fixed-effect model embedded in the METASOFT tool\u003csup\u003e\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e\u003c/sup\u003e (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://genetics.cs.ucla.edu/meta/\u003c/span\u003e\u003cspan address=\"http://genetics.cs.ucla.edu/meta/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) was used to perform trans-ancestral GWAS meta-analyses between EAS and EUR for 3,414 brain imaging phenotypes that were generated in similar ways from CHIMGEN and UKBB. GWAS summary statistics of the discovery stages of CHIMGEN (5,025 participants) and UKBB (22,138 participants) were used to conduct trans-ancestral GWAS meta-analyses in the discovery stage, while GWAS summary statistics of the replication stages of CHIMGEN (2,033 participants) and UKBB (11,086 participants) were used to perform trans-ancestral GWAS meta-analyses in the replication stage. The same approach as the EAS-GWASs was used to identify independent variant-phenotype associations in the trans-ancestral GWAS meta-analyses with different statistical thresholds and multiple testing correction methods in the discovery (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;5 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;8\u003c/sup\u003e and \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;1.46 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;11\u003c/sup\u003e) and replication (uncorrected, FDR-corrected, and Bonferroni-corrected \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05) samples.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec19\" class=\"Section2\"\u003e \u003ch2\u003eIdentifying novel associations from EAS-GWASs and trans-ancestral GWASs\u003c/h2\u003e \u003cp\u003eTo identify novel variant-phenotype associations from EAS-GWASs and trans-ancestral GWASs, we defined the known independent significant associations as those with \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;1.46 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;11\u003c/sup\u003e in the discovery stage (22,138 participants) and \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05 in the replication stage (11,086 participants) in the currently largest EUR-GWASs for the 3,414 brain imaging phenotypes\u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e. With the same thresholds (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;1.46 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;11\u003c/sup\u003e for discovery and \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05 for replication), we created the lists of independent variant-phenotype associations for EAS-GWASs and trans-ancestral GWASs. In EAS-GWASs, a novel variant-phenotype association was defined as the corresponding locus of the variant that did not overlap with any known loci for the same brain imaging phenotype in EUR-GWASs. In trans-ancestral GWAS meta-analyses, a novel variant-phenotype association was defined as the corresponding locus of the variant that did not overlap with any loci of the same brain imaging phenotype in either EUR-GWASs or EAS-GWASs.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec20\" class=\"Section2\"\u003e \u003ch2\u003ePooled genome-wide significant associations\u003c/h2\u003e \u003cp\u003eWe pooled all genome-wide significant associations (discovered at \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;5 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;8\u003c/sup\u003e and replicated at \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05) identified in any of the EAS-GWASs, EUR-GWASs, or trans-ancestral GWASs for each of the 3,414 brain imaging phenotypes. Specifically, for each phenotype, the overlapping associated loci among these three types of GWASs were merged into an independent locus indexed by the most significant variant of these loci.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec21\" class=\"Section2\"\u003e \u003ch2\u003eEnrichment analysis for the pooled genome-wide significant associations\u003c/h2\u003e \u003cp\u003ePANTHER (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://pantherdb.org/\u003c/span\u003e\u003cspan address=\"http://pantherdb.org/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) was used to conduct functional enrichment analysis. We first assigned the index variant of each pooled genome-wide significant association to a protein-coding gene (gencode.v38lift37.annotation.gtf.gz) if the variant located within 10 kb around the gene. Then the obtained genes were used to test functional enrichment in pathways derived from Reactome Pathway Database using all protein-coding genes (n\u0026thinsp;=\u0026thinsp;20,589) as the background. Significance of functional enrichment was calculated using Fisher\u0026rsquo;s exact test with an FDR-corrected \u003cem\u003eq\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec22\" class=\"Section2\"\u003e \u003ch2\u003eAncestry-shared and ancestry-specific associations between EAS and EUR\u003c/h2\u003e \u003cp\u003eThe Cochran\u0026rsquo;s Q-test (CQ-test) was applied to examine the effect size differences in variant-phenotype associations between EAS and EUR. The pooled independent significant associations (discovered at \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;5 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;8\u003c/sup\u003e and replicated at \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05) whose index variants studied in both EAS- and EUR-GWASs were included in the heterogeneous assessment with a discovery-replication scheme. In the discovery stage, the CQ-test was conducted based on GWAS summary statistics from the discovery samples of CHIMGEN (n\u0026thinsp;=\u0026thinsp;5,025) and UKBB (n\u0026thinsp;=\u0026thinsp;22,138). In the replication stage, the CQ-test was performed based on GWAS summary statistics from the replication samples of CHIMGEN (n\u0026thinsp;=\u0026thinsp;2,033) and UKBB (n\u0026thinsp;=\u0026thinsp;11,086).\u003c/p\u003e \u003cp\u003eBased on the heterogeneous assessment, the variant-phenotype associations were divided into three categories of ancestry-shared, ancestry-specific, and un-classified associations. The ancestry-shared associations were defined as those variant-phenotype associations with \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026ge;\u0026thinsp;0.05 in both discovery and replication CQ-tests. The ancestry-specific associations were defined as those variant-phenotype associations with \u003cem\u003ePc\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05 (Bonferroni correction for the number of the included associations) in the discovery CQ-tests and uncorrected, FDR-corrected, and Bonferroni-corrected \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05 in the replication CQ-tests.\u003c/p\u003e \u003cp\u003eSince the unbalanced sample sizes between EAS (n\u0026thinsp;=\u0026thinsp;5,025 for discovery and n\u0026thinsp;=\u0026thinsp;2,033 for replication) and EUR (n\u0026thinsp;=\u0026thinsp;22,138 for discovery and n\u0026thinsp;=\u0026thinsp;11,086 for replication) may bring bias to the heterogeneous assessment, we also validate the identified ancestry-shared and ancestry-specific associations in the EAS (n\u0026thinsp;=\u0026thinsp;7,058 from CHIMGEN) and EUR (n\u0026thinsp;=\u0026thinsp;8,428 from UKBB)\u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e populations with comparable sample size. Because only 2,640 of the 3,414 brain imaging phenotypes were included in the EUR-GWASs\u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e, we filtered out the identified ancestry-shared and ancestry-specific associations that were absent in the EUR-GWASs\u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e. Therefore, the CQ-tests were only performed for the remaining ancestry-shared associations and ancestry-specific associations. The ancestry-shared associations were deemed to be validated when \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026ge;\u0026thinsp;0.05. We reported the ancestry-specific associations that were validated at uncorrected, FDR-corrected, and Bonferroni-corrected \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec23\" class=\"Section2\"\u003e \u003ch2\u003eGenetic colocalizations between brain imaging phenotypes and brain-related non-imaging traits\u003c/h2\u003e \u003cp\u003eWe tested whether brain imaging phenotypes had shared genetic architectures with various brain-related non-imaging traits. We only included non-imaging traits associated with brain structure and function and with available GWAS summary statistics. If more than one GWASs were available for a trait, we only included the GWAS summary statistics with the largest number of independent associations. For non-imaging traits with available GWAS summary statistics for both EUR and EAS populations, we performed trans-ancestral GWAS meta-analyses with fixed-effect model to integrate the results. The finally included 37 non-imaging traits consisted of 3 traits for cognition (cognitive performance, educational attainment, and intelligence); 13 for personality and behavior (aggressive behavior, antisocial behavior, depressive symptoms, subjective well-being, agreeableness, conscientiousness, openness, extraversion, neuroticism, risky behaviors, risk tolerance, insomnia symptoms and isolation); 2 for addiction (drinks per week and smoking initiation); 11 for psychiatric disorders (anorexia nervosa, anxiety disorder, attention deficit hyperactivity disorder, autism spectrum disorder, bipolar disorder, major depressive disorder, obsessive compulsive disorder, panic disorder, post-traumatic stress disorder, schizophrenia, and Tourette syndrome); and 8 for neurological disorders (Alzheimer\u0026rsquo;s disease, amyotrophic lateral sclerosis, any stroke, focal epilepsy, genetic generalized epilepsy, migraine, multiple sclerosis, and Parkinson\u0026rsquo;s disease) (Supplementary Table\u0026nbsp;12). We generated independent significant loci (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;5 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;8\u003c/sup\u003e) for each non-imaging trait using an iterative process including the following procedures: (a) identifying the most significant variant; (b) grouping all variants within 500 kb centered at the variant into a locus; and (c) merging overlapping loci.\u003c/p\u003e \u003cp\u003eHere, we used three continuous steps to identify shared genetic architectures between brain-related non-imaging traits and brain imaging phenotypes. In step 1, we screened non-imaging traits with at least one locus containing an index variant of the pooled genome-wide significant associations for brain imaging phenotypes (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;5 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;8\u003c/sup\u003e for discovery and \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05 for replication). In step 2, we split the autosome into 11,508 chunks with 500kb (similar with the mean size of the significant loci for the non-imaging traits) and defined the probability of a chunk that contained at least one index variant for brain imaging phenotypes as the reference probability of non-significance. For each non-imaging trait survived in step 1, Fisher\u0026rsquo;s exact test was performed to test whether the colocalization probability of the trait with brain imaging phenotypes was significantly higher than the reference probability. In step 3, we used vSampler\u003csup\u003e\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e\u003c/sup\u003e to randomly generate 1000 sets of variants with matched allele frequency, gene proximity, and number of LD proxies with the index variants of brain imaging phenotypes using the EUR and EAS from 1KGP as reference panels, respectively. For each non-imaging trait also survived in step 2, we tested whether the number of colocalized trait-related loci with the index variants of brain imaging phenotypes was significantly higher than the number of colocalized trait-related loci with randomly generated variants using the resampling strategy (1,000 sets, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05). The non-imaging traits survived in the three continuous tests were deemed to have shared genetic architectures with brain imaging phenotypes. To determine the specific loci shared by each pair of imaging and non-imaging traits, we only included the loci of the non-imaging trait whose index variants showed a high LD (r2\u0026thinsp;\u0026gt;\u0026thinsp;0.8) with the index variants of the brain imaging phenotype.\u003c/p\u003e \u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe are grateful to participants and researchers of CHIMGEN, who generously donated their time to make this resource available. We acknowledge funding from the National Key Research and Development Program of China (2018YFC1314300) to Chunshui Yu, and the National Natural Science Foundation of China (82030053, 81425013) to Chunshui Yu. We thank SG10K Consortium for collecting and sharing the high-coverage whole genome sequencing data of Asian Populations. We acknowledge UK Biobank for providing GWAS summary statistics of brain imaging phenotypes. For the genetic colocalization analyses, we used summary statistical data from several GWASs of brain-related non-imaging traits. We thank groups [BioBank Japan (BBJ); the Complex Traits Genetics laboratory (CTGlab); the Early Genetics and Lifecourse Epidemiology consortium (EAGLE); Genetics of Personality Consortium (GPC); the GWAS and Sequencing Consortium of Alcohol and Nicotine use (GSCAN); the International Headache Genetics Consortium (IHGC); the International League Against Epilepsy (ILAE); International Multiple Sclerosis Genetics Consortium (IMSGC); METASTROKE collaboration, the Psychiatric Genomics Consortium (PGC); the Social Science Genetic Association Consortium (SSGAC); UK Biobank (UKBB)], authors (Jacqueline M Lane, Jia Nee Foo, Mike A Nalls and Wouter van Rheenen) for making these data publicly available and all the participants and researchers in these studies.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eChunshui Yu and Jilian Fu designed the study and wrote the manuscript. Jilian Fu, Jianhua Wang and Quan Zhang analyzed the data. Chunshui Yu, Mulin Jun Li and Jingliang Cheng supervised this work. Jilian Fu, Quan Zhang, Jianhua Wang, Meiyun Wang, Bing Zhang, Wenzhen Zhu, Shijun Qiu,\u0026nbsp;Zuojun Geng, Guangbin Cui, Yongqiang Yu, Weihua Liao, Hui Zhang, Bo Gao, Xiaojun Xu, Tong Han, Zhenwei Yao, Wen Qin, Feng Liu, Meng Liang, Sijia Wang, Qiang Xu, Jiayuan Xu, Peng Zhang, Wei Li, Dapeng Shi, Caihong Wang, Su Lui, Zhihan Yan, Feng Chen, Jing Zhang, Jiance Li, Wen Shen, Yanwei Miao, Dawei Wang, Junfang Xian, Jia-Hong Gao, Xiaochu Zhang, Kai Xu, Xi-Nian Zuo, Longjiang Zhang, Zhaoxiang Ye, Jingliang Cheng, Mulin Jun Li and Chunshui Yu acquired the data. All authors critically reviewed the manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCode availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll the software and code used in this study are publicly available\u003c/p\u003e\n\u003cp\u003eBGENEv1.2 (https://jmarchini.org/bgenie/)\u003c/p\u003e\n\u003cp\u003eCAT12 (http://dbm.neuro.uni-jena.de/cat)\u003c/p\u003e\n\u003cp\u003eComBat harmonization (https://github.com/precision-medicine-um/ComBatHarmonization)\u003c/p\u003e\n\u003cp\u003eFMRIB\u0026rsquo;s Integrated Registration and Segmentation Tool (FIRST) (https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FIRST)\u003c/p\u003e\n\u003cp\u003eFMRIB Software Library (FSL, version 5.0.10; http://www.fmrib.ox.ac.uk/fsl)\u003c/p\u003e\n\u003cp\u003eFreeSurfer v6.0.0 (http://surfer.nmr.mgh.harvard.edu/)\u003c/p\u003e\n\u003cp\u003eIMPUTE2 (http://mathgen.stats.ox.ac.uk/impute/impute_v2.html)\u003c/p\u003e\n\u003cp\u003eKING (https://www.kingrelatedness.com/)\u003c/p\u003e\n\u003cp\u003eMETASOFT (http://genetics.cs.ucla.edu/meta/)\u003c/p\u003e\n\u003cp\u003ePANTHER (http://pantherdb.org/)\u003c/p\u003e\n\u003cp\u003ePheWeb (https://github.com/statgen/pheweb/)\u003c/p\u003e\n\u003cp\u003ePLINKv2.0 (http://zzz.bwh.harvard.edu/plink/)\u003c/p\u003e\n\u003cp\u003eSHAPEIT2 (https://mathgen.stats.ox.ac.uk/genetics_software/shapeit/shapeit.html)\u003c/p\u003e\n\u003cp\u003eSPM12 (\u003ca href=\"https://www.fil.ion.ucl.ac.uk/spm/software/spm12/\"\u003ehttps://www.fil.ion.ucl.ac.uk/spm/software/spm12/\u003c/a\u003e)\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll GWAS results based on Chinese Han participants are available on the website of CHIMGEN (http://chimgen.tmu.edu.cn/pheweb/), which allows users to browse associations by variants, genes, and brain imaging phenotypes. All GWAS results based on UK Biobank participants are available on the website of Oxford Brain Imaging Genetics (BIG) web browser (http://big.stats.ox.ac.uk/). 1000 Genomes Project reference panel can be found on https://mathgen.stats.ox.ac.uk/impute/1000GP_Phase3.html. And the high-coverage whole genome sequencing data of SG10K is available on https://ega-archive.org with accession number of EGAS00001003875.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAll authors of CHIMGEN Consortium\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDepartment of Radiology and Tianjin Key Laboratory of Functional Imaging, Tianjin Medical University General Hospital\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eChunshui Yu, Quan Zhang, Wen Qin, Feng Liu, Junping Wang, Qiang Xu, Jiayuan Xu, Xue Zhang, Xinjun Suo, Jilian Fu, Congcong Yuan, Yuan Ji, Hui Xue, Tianying Gao, Junpeng Liu, Yanjun Li, Xi Guo, Lixue Xu, Jiajia Zhu, Huaigui Liu, Fangshi Zhao, Jie Sun, Yongjie Xu, Huanhuan Cai, Jie Tang, Yaodan Zhang, Yongqin Xiong, Xianting Sun, Nannan Pan, Xue Zhang (Junior), Jiayang Yang, Nana Liu, Ya Wen, Dan Zhu, Bingjie Wu, Wenshuang Zhu, Qingqing Diao, Yujuan Cao, Bingbing Yang, Lining Guo, Yingying Xie, Jiahui Lin, Zhimin Li, Yan Zhang, Kaizhong Xue, Zirui Wang, Junlin Shen\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSchool of Medical Imaging, Tianjin Medical University\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eMeng Liang, Xuejun Zhang, Hao Ding, Qian Su, Sijia Wang\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDepartment of Bioinformatics, The Province and Ministry Co-sponsored Collaborative Innovation Center for Medical Epigenetics, School of Basic Medical Sciences, Tianjin Medical University\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eMulin Jun Li, Shijie Zhang, Jianhua Wang\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDepartment of Radiology, Tianjin Medical University Cancer Institute and Hospital\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eZhaoxiang Ye, Peng Zhang, Wei Li\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDepartment of Radiology, Henan Provincial People\u0026rsquo;s Hospital \u0026amp; Zhengzhou University People\u0026rsquo;s Hospital\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eMeiyun Wang, Dapeng Shi, Lun Ma, Yan Bai, Min Guan, Wei Wei\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDepartment of Magnetic Resonance Imaging, The First Affiliated Hospital of Zhengzhou University\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eJingliang Cheng, Caihong Wang, Peifang Miao, Fuhong Duan, Yafei Guo, Weijian Wang\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDepartment of Medical Imaging, Jinling Hospital, Medical School of Nanjing University\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eLongjiang Zhang, Lijuan Zheng, Li Lin, Yunfei Wang, Han Zhang, Xinyuan Zhang\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDepartment of Radiology, Drum Tower Hospital, Medical School of Nanjing University\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eBing Zhang, Zhao Qing, Sichu Wu, Junxia Wang, Yi Sun, Yang He\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eInstitute of Psychology, Chinese Academy of Sciences\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eXi-Nian Zuo, Zhe Zhang, Yin-Shan Wang, Quan Zhou\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDepartment of Radiology, Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWenzhen Zhu, Tian Tian\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDepartment of Medical Imaging, The First Affiliated Hospital of Guangzhou University of Traditional Chinese Medicine\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eShijun Qiu, Yi Liang, Yujie Liu, Hui Zeng, Jingxian Chen\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDepartment of Radiology, The Affiliated Hospital of Xuzhou Medical University\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eKai Xu, Haitao Ge, Peng Xu, Cailuan Lu, Chen Wu, Xiaoying Yang\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDepartment of Medical Imaging, The Second Hospital of Hebei Medical University\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eZuojun Geng, Yuzhao Wang, Yankai Wu, Xuran Feng, Ling Li, Duo Gao\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDivision of Life Science and Medicine, University of Science \u0026amp; Technology of China\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eXiaochu Zhang, Rujing Zha, Ying Li, Lizhuang Yang, Ying Chen, Ling Zuo\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCenter for MRI Research, Academy for Advanced Interdisciplinary Studies, Peking University\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eJia-Hong Gao, Jianqiao Ge, Guoyuan Yang\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunctional and Molecular Imaging Key Lab of Shaanxi Province \u0026amp; Department of Radiology, Tangdu Hospital, Air Force Medical University\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eGuangbin Cui, Wen Wang, Linfeng Yan, Yang Yang, Jin Zhang\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDepartment of Radiology, Beijing Tongren Hospital, Capital Medical University\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eJunfang Xian, Qian Wang, Xiaoxia Qu, Ying Wang\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDepartment of Radiology, Characteristic Medical Center of Chinese People\u0026rsquo;s Armed Police Force\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eQuan Zhang, Fei Yuan\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDepartment of Radiology, Qilu Hospital of Shandong University\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eDawei Wang, Li Hu, Jizhen Li\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDepartment of Radiology, The First Affiliated Hospital of Dalian Medical University\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eYanwei Miao, Weiwei Wang, Yujing Zhou\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDepartment of Radiology, Tianjin First Center Hospital\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWen Shen, Miaomiao Long, Lihua Liu\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDepartment of Radiology, The First Affiliated Hospital of Anhui Medical University\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eYongqiang Yu, Xiaohu Li, Xiaoshu Li\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDepartment of Radiology, The First Affiliated Hospital of Wenzhou Medical University\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eJiance Li, Yunjun Yang, Nengzhi Xia\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDepartment of Radiology, Xiangya Hospital, Central South University\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWeihua Liao, Shuai Yang, Youming Zhang\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDepartment of Magnetic Resonance, Lanzhou University Second Hospital\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eJing Zhang, Guangyao Liu, Laiyang Ma\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDepartment of Radiology, The First Hospital of Shanxi Medical University\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eHui Zhang, Xiaochun Wang, Ying Lei\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDepartment of Radiology, Yantai Yuhuangding Hospital\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eBo Gao, Gang Zhang, Kang Yuan\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDepartment of Radiology, The Second Affiliated Hospital of Zhejiang University, School of Medicine\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eXiaojun Xu, Jingjing Xu, Xiaojun Guan\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDepartment of Radiology, Hainan General Hospital\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFeng Chen, Yuankai Lin, Huijuan Chen\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDepartment of Radiology, The Second Affiliated Hospital and Yuying Children\u0026rsquo;s Hospital of Wenzhou Medical University\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eZhihan Yan, Yuchuan Fu, Yi Lu\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDepartment of Radiology, Tianjin Huanhu Hospital\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTong Han, Jun Guo, Hao Lu\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDepartment of Radiology, Huashan Hospital, Fudan University\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eZhenwei Yao, Yue Wu\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDepartment of Radiology, the Center for Medical Imaging, West China Hospital of Sichuan University\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSu Lui\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eDeco, G. \u0026amp; Kringelbach, M. L. Great Expectations: Using Whole-Brain Computational Connectomics for Understanding Neuropsychiatric Disorders. Neuron \u003cb\u003e84\u003c/b\u003e, 892\u0026ndash;905 (2014).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJansen, A. G., Mous, S. E., White, T., Posthuma, D. \u0026amp; Polderman, T. J. C. What Twin Studies Tell Us About the Heritability of Brain Development, Morphology, and Function: A Review. Neuropsychol Rev \u003cb\u003e25\u003c/b\u003e, 27\u0026ndash;46 (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMakowski, C. et al. Discovery of genomic loci of the human cerebral cortex using genetically informed brain atlases. Science \u003cb\u003e375\u003c/b\u003e, 522\u0026ndash;528 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eElliott, L. T. et al. Genome-wide association studies of brain imaging phenotypes in UK Biobank. Nature \u003cb\u003e562\u003c/b\u003e, 210\u0026ndash;216 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhao, B. X. et al. Common genetic variation influencing human white matter microstructure. Science \u003cb\u003e372\u003c/b\u003e, 1304\u0026ndash;1316 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHibar, D. P. et al. Common genetic variants influence human subcortical brain structures. Nature \u003cb\u003e520\u003c/b\u003e, 224\u0026ndash;229 (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGrasby, K. L. et al. The genetic architecture of the human cerebral cortex. Science \u003cb\u003e367\u003c/b\u003e (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKang, D. W. et al. Differences in cortical structure between cognitively normal East Asian and Caucasian older adults: a surface-based morphometry study. Scientific reports \u003cb\u003e10\u003c/b\u003e (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTang, Y. C. et al. Brain structure differences between Chinese and Caucasian cohorts: A comprehensive morphometry study. Human brain mapping \u003cb\u003e39\u003c/b\u003e, 2147\u0026ndash;2155 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTang, Y. C. et al. The construction of a Chinese MRI brain atlas: A morphometric comparison study between Chinese and Caucasian cohorts. NeuroImage \u003cb\u003e51\u003c/b\u003e, 33\u0026ndash;41 (2010).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLanger-Gould, A., Brara, S. M., Beaber, B. E. \u0026amp; Zhang, J. L. Incidence of multiple sclerosis in multiple racial and ethnic groups. Neurology \u003cb\u003e80\u003c/b\u003e, 1734\u0026ndash;1739 (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePapp, V. et al. Worldwide Incidence and Prevalence of Neuromyelitis Optica A Systematic Review. Neurology \u003cb\u003e96\u003c/b\u003e, 59\u0026ndash;77 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMeeker, K. L. et al. Socioeconomic Status Mediates Racial Differences Seen Using the AT(N) Framework. Annals of neurology \u003cb\u003e89\u003c/b\u003e, 254\u0026ndash;265 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePeterson, R. E. et al. Genome-wide Association Studies in Ancestrally Diverse Populations: Opportunities, Methods, Pitfalls, and Recommendations. Cell \u003cb\u003e179\u003c/b\u003e, 589\u0026ndash;603 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSmith, S. M. et al. An expanded set of genome-wide association studies of brain imaging phenotypes in UK Biobank. Nature neuroscience \u003cb\u003e24\u003c/b\u003e, 737\u0026ndash;745 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGraham, S. E. et al. The power of genetic diversity in genome-wide association studies of lipids. Nature (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen, M. H. et al. Trans-ethnic and Ancestry-Specific Blood-Cell Genetics in 746,667 Individuals from 5 Global Populations. Cell \u003cb\u003e182\u003c/b\u003e, 1198\u0026ndash;1213 e1114 (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXu, Q. et al. CHIMGEN: a Chinese imaging genetics cohort to enhance cross-ethnic and cross-geographic brain research. Molecular psychiatry \u003cb\u003e25\u003c/b\u003e, 517\u0026ndash;529 (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMela, A. \u0026amp; Goldman, J. E. The Tetraspanin KAI1/CD82 Is Expressed by Late-Lineage Oligodendrocyte Precursors and May Function to Restrict Precursor Migration and Promote Oligodendrocyte Differentiation and Myelination. Journal of Neuroscience \u003cb\u003e29\u003c/b\u003e, 11172\u0026ndash;11181 (2009).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhao, Y. et al. CD82-TRPM7-Numb signaling mediates age-related cognitive impairment. Geroscience \u003cb\u003e42\u003c/b\u003e, 595\u0026ndash;611 (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSuzuki, T. et al. Genome-wide association study of epilepsy in a Japanese population identified an associated region at chromosome 12q24. Epilepsia \u003cb\u003e62\u003c/b\u003e, 1391\u0026ndash;1400 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHautakangas, H. et al. Genome-wide analysis of 102,084 migraine cases identifies 123 risk loci and subtype-specific risk alleles. Nature genetics \u003cb\u003e54\u003c/b\u003e, 152\u0026ndash;160 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHong, S. E. et al. Autosomal recessive lissencephaly with cerebellar hypoplasia is associated with human RELN mutations. Nature genetics \u003cb\u003e26\u003c/b\u003e, 93\u0026ndash;96 (2000).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMao, D. X. et al. De novo EIF2AK1 and EIF2AK2 Variants Are Associated with Developmental Delay, Leukoencephalopathy, and Neurologic Decompensation. American journal of human genetics \u003cb\u003e106\u003c/b\u003e, 570\u0026ndash;583 (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen, H. R., Juan, H. C., Wong, Y. H., Tsai, J. W. \u0026amp; Fann, M. J. Cdk12 Regulates Neurogenesis and Late-Arising Neuronal Migration in the Developing Cerebral Cortex. Cerebral cortex \u003cb\u003e27\u003c/b\u003e, 2289\u0026ndash;2302 (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGreenbaum, L. \u0026amp; Lerer, B. Differential contribution of genetic variation in multiple brain nicotinic cholinergic receptors to nicotine dependence: recent progress and emerging open questions. Molecular psychiatry \u003cb\u003e14\u003c/b\u003e, 912\u0026ndash;945 (2009).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWight, T. N. Versican: a versatile extracellular matrix proteoglycan in cell biology. Current opinion in cell biology \u003cb\u003e14\u003c/b\u003e, 617\u0026ndash;623 (2002).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZang, Z. S., Xu, Y. M. \u0026amp; Lau, A. T. Y. Molecular and pathophysiological aspects of metal ion uptake by the zinc transporter ZIP8 (SLC39A8). Toxicol Res-Uk \u003cb\u003e5\u003c/b\u003e, 987\u0026ndash;1002 (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSong, W. C., Qian, W., Wang, W. D., Yu, S. Y. \u0026amp; Lin, G. N. Mendelian randomization studies of brain MRI yield insights into the pathogenesis of neuropsychiatric disorders. BMC genomics \u003cb\u003e22\u003c/b\u003e (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBuch, A. M. \u0026amp; Liston, C. Dissecting diagnostic heterogeneity in depression by integrating neuroimaging and genetics. Neuropsychopharmacology: official publication of the American College of Neuropsychopharmacology \u003cb\u003e46\u003c/b\u003e, 156\u0026ndash;175 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAbi-Dargham, A. \u0026amp; Horga, G. The search for imaging biomarkers in psychiatric disorders. Nat Med \u003cb\u003e22\u003c/b\u003e, 1248\u0026ndash;1255 (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBycroft, C. et al. The UK Biobank resource with deep phenotyping and genomic data. Nature \u003cb\u003e562\u003c/b\u003e, 203\u0026ndash;209 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRajapakse, J. C., Giedd, J. N. \u0026amp; Rapoport, J. L. Statistical approach to segmentation of single-channel cerebral MR images. IEEE transactions on medical imaging \u003cb\u003e16\u003c/b\u003e, 176\u0026ndash;186 (1997).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAshburner, J. A fast diffeomorphic image registration algorithm. NeuroImage \u003cb\u003e38\u003c/b\u003e, 95\u0026ndash;113 (2007).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSmith, S. M. et al. Tract-based spatial statistics: Voxelwise analysis of multi-subject diffusion data. NeuroImage \u003cb\u003e31\u003c/b\u003e, 1487\u0026ndash;1505 (2006).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJenkinson, M., Bannister, P., Brady, M. \u0026amp; Smith, S. Improved optimization for the robust and accurate linear registration and motion correction of brain images. NeuroImage \u003cb\u003e17\u003c/b\u003e, 825\u0026ndash;841 (2002).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFortin, J. P. et al. Harmonization of multi-site diffusion tensor imaging data. NeuroImage \u003cb\u003e161\u003c/b\u003e, 149\u0026ndash;170 (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlfaro-Almagro, F. et al. Image processing and Quality Control for the first 10,000 brain imaging datasets from UK Biobank. NeuroImage \u003cb\u003e166\u003c/b\u003e, 400\u0026ndash;424 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePurcell, S. et al. PLINK: a tool set for whole-genome association and population-based linkage analyses. American journal of human genetics \u003cb\u003e81\u003c/b\u003e, 559\u0026ndash;575 (2007).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAnderson, C. A. et al. Data quality control in genetic case-control association studies. Nat Protoc \u003cb\u003e5\u003c/b\u003e, 1564\u0026ndash;1573 (2010).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWu, D. et al. Large-Scale Whole-Genome Sequencing of Three Diverse Asian Populations in Singapore. Cell \u003cb\u003e179\u003c/b\u003e, 736\u0026ndash;749 e715 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDelaneau, O., Zagury, J. F. \u0026amp; Marchini, J. Improved whole-chromosome phasing for disease and population genetic studies. Nature methods \u003cb\u003e10\u003c/b\u003e, 5\u0026ndash;6 (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHowie, B., Fuchsberger, C., Stephens, M., Marchini, J. \u0026amp; Abecasis, G. R. Fast and accurate genotype imputation in genome-wide association studies through pre-phasing. Nature genetics \u003cb\u003e44\u003c/b\u003e, 955\u0026ndash;959 (2012).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eManichaikul, A. et al. Robust relationship inference in genome-wide association studies. Bioinformatics \u003cb\u003e26\u003c/b\u003e, 2867\u0026ndash;2873 (2010).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLan, T. M. et al. Deep whole-genome sequencing of 90 Han Chinese genomes. GigaScience \u003cb\u003e6\u003c/b\u003e (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHan, B. \u0026amp; Eskin, E. Random-Effects Model Aimed at Discovering Associations in Meta-Analysis of Genome-wide Association Studies. American journal of human genetics \u003cb\u003e88\u003c/b\u003e, 586\u0026ndash;598 (2011).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHuang, D. D. et al. vSampler: fast and annotation-based matched variant sampling tool. Bioinformatics \u003cb\u003e37\u003c/b\u003e, 1915\u0026ndash;1917 (2021).\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"nature-portfolio","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"","title":"Nature Portfolio","twitterHandle":"","acdcEnabled":false,"dfaEnabled":false,"editorialSystem":"ejp","reportingPortfolio":"","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-2047527/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-2047527/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eGenome-wide association studies of brain imaging phenotypes are mainly performed in European populations, but other populations are severely under-represented. Here, we conducted Chinese-alone and trans-ancestral genome-wide association studies of 3,414 brain imaging phenotypes in 7,058 Chinese and 33,224 European individuals. We identified 37 novel variant-phenotype associations in Chinese-alone analyses and 459 additional novel associations in trans-ancestral meta-analyses under the thresholds of \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;1.46 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;11\u003c/sup\u003e for discovery and \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05 for replication. We pooled genome-wide significant associations for brain imaging phenotypes identified in either single-ancestral or trans-ancestral analyses into 6,361 independent significant associations. These associations were unevenly distributed in the genome and across the brain phenotypic subgroups and demonstrated significant enrichment for nervous system development and signal transduction. We further categorized the 4,890 pooled genome-wide significant associations whose index variants were included in both Chinese and European analyses into 43 ancestry-specific and 3,524 ancestry-shared associations. Loci of the 6,361 pooled genome-wide significant associations for brain imaging phenotypes were shared by 16 brain-related non-imaging traits including cognition, personality, risk behavior, addiction, and neuropsychiatric disorders. Our results provide a valuable catalog of genetic associations for brain imaging phenotypes in diverse populations.\u003c/p\u003e","manuscriptTitle":"Trans-ancestral genome-wide association studies of brain imaging phenotypes","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2022-09-26 13:40:34","doi":"10.21203/rs.3.rs-2047527/v1","editorialEvents":[],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"nature-genetics","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"ng","sideBox":"Learn more about [Nature Genetics](http://www.nature.com/ng/)","snPcode":"","submissionUrl":"","title":"Nature Genetics","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Nature Research","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"11ff3e29-8076-408f-972a-ba16b8a6459d","owner":[],"postedDate":"September 26th, 2022","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2024-05-30T07:06:34+00:00","versionOfRecord":{"articleIdentity":"rs-2047527","link":"https://doi.org/10.1038/s41588-024-01766-y","journal":{"identity":"nature-genetics","isVorOnly":false,"title":"Nature Genetics"},"publishedOn":"2024-05-29 04:00:00","publishedOnDateReadable":"May 29th, 2024"},"versionCreatedAt":"2022-09-26 13:40:34","video":"","vorDoi":"10.1038/s41588-024-01766-y","vorDoiUrl":"https://doi.org/10.1038/s41588-024-01766-y","workflowStages":[]},"version":"v1","identity":"rs-2047527","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-2047527","identity":"rs-2047527","version":["v1"]},"buildId":"rHA-KDH7Qsr4HCuvH75dn","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}