Genetic Architecture of Schizophrenia Clinical Subtypes

Genetic Architecture of Schizophrenia Clinical Subtypes | 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 Research Article Genetic Architecture of Schizophrenia Clinical Subtypes Meng Zhou, Yamin Zhang, Yilu Zhao, Zheng Fu, Xiaohui Wang, Xinglun Dang, and 9 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8620966/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 10 You are reading this latest preprint version Abstract Background Clinical heterogeneity in schizophrenia (SZ) presents a significant challenge to genetic research, as diverse symptom profiles may stem from distinct genetic risk factors. Although previous studies have stratified patients into symptom-based subtypes, and preliminary evidence suggests the presence of distinct architectures, these findings remain limited. This study aimed to identify the clinically defined SZ subtypes and investigate the genetic architecture underlying different subtypes. Methods In a Chinese Han cohort of 2,410 SZ patients, we applied K-means cluster analysis to symptom profiles to identify clinical subtypes. The identified subtype structure was validated in an independent cohort of 480 patients. Subsequently, subtype-specific genome-wide association studies (GWAS) were conducted to identify genetic risk loci associated with individual subtypes. Results Three stable subtypes were identified: Cluster-L (low severity), Cluster-S (severe), and Cluster-N (predominant negative symptoms). Reproducibility of this classification was confirmed in the independent cohort. The three subtypes also exhibited significantly different symptom network structures. In GWAS analysis, A total of four genome-wide significant loci were detected, including a Cluster-N–specific locus within the CNTN2 gene (lead SNP rs3767295, P = 5.30E-10, OR = 0.62). Gene-based analyses revealed additional risk genes unique to particular subtypes. Moreover, subtype-specific patterns emerged in both pathway-specific polygenic risk scores (pPRS) and cell type–specific PRS (ctPRS). Conclusions These findings underscore the value of patient stratification in improving statistical power to detect subtype-specific risk loci. They further demonstrate that SZ patients with distinct symptom profiles harbor differential genetic liabilities involving diverse biological pathways and cell types. Schizophrenia Clinical subtypes Genetics Heterogeneity Figures Figure 1 Figure 2 Figure 3 Figure 4 Background Schizophrenia (SZ) is a highly heritable psychotic disorder characterized by both complex genetic architecture and diverse clinical manifestations( 1 ). Large-scale genome-wide association studies (GWAS) have identified more than 200 risk loci associated with SZ( 2 ). However, these findings explain only approximately 40% of the estimated heritability, leaving a substantial proportion unaccounted for( 3 ). A major barrier to resolving this “missing heritability” is the marked heterogeneity in both clinical presentation and genetic architecture. Distinct symptom profiles and treatment responses may reflect different genetic liabilities( 4 ). For example, severe negative and disorganized symptoms have been linked to stronger family histories of psychiatric illness( 5 ) and higher SZ polygenic risk scores (PRS)( 6 ). Elevated PRS have also been associated with treatment resistance( 7 ) and earlier onset( 8 ). These findings underscore the limitations of treating SZ as a single, uniform entity in genetic analyses. Subtyping patients into more clinically homogeneous groups offers a promising avenue for genetic discovery( 9 ). Several studies have stratified patients based on symptom dimensions, cognitive impairments, or treatment response, identifying subgroups with distinct genetic profiles( 10 – 12 ). For instance, variants in the chromosome 1q region showed significant associations with the negative symptom subtype but not when all patients were analyzed together( 11 ). Beyond SZ, phenotype-driven stratification has also revealed distinct liabilities across the broader psychosis spectrum( 7 , 13 ). Collectively, these findings suggest that identifying clinically meaningful subgroups could help bridge the gap in unexplained heritability, clarify biological mechanisms, and ultimately inform more precise therapeutic strategies for SZ. Despite increasing interest in the genetic underpinnings of SZ subtypes, most existing research have focused on genome-wide PRS (gPRS) to estimate overall liability. While informative, such approaches provide limited insight into the specific biological processes that differentiate subtypes. In contrast, subtype-specific GWAS have proven powerful in other psychiatric disorders, including bipolar disorder( 14 ), major depressive disorder( 15 ), and autism( 16 ), where they have revealed subgroup-specific genetic architectures that would otherwise be masked in aggregated analyses. In SZ, however, subtype-based GWAS remain relatively rare. Expanding the application of subtype-focused genetic analyses in SZ is therefore crucial for uncovering novel risk loci and clarifying disease mechanisms. In this study, we sought to address this gap by investigating the genetic heterogeneity underlying clinically defined subtypes of SZ. We first classified patients into distinct clinical subtypes based on their symptom profiles and examined their symptom network structures. We then conducted subtype-specific GWAS and gene-based analyses to identify loci and genes associated with particular subgroups. Finally, we evaluated whether genetic risks, both genome-wide and within specific biological pathways and brain cell types, varied across subtypes. Through this approach, we aimed to demonstrate that stratification by clinical subtype can enhance the discovery of biologically meaningful genetic signals that are obscured in traditional case–control GWAS. Methods Study subjects A total of 2,896 patients with SZ were recruited from eight psychiatric hospitals across Sichuan and Zhejiang provinces in China, along with 3,000 healthy controls recruited from the community. Recruitment occurred in two phases. The discovery cohort comprised 2,410 patients enrolled between 2021 and 2022, while an additional 480 patients were recruited between 2023 and 2025 for replication. All participants were of Han Chinese ancestry, and detailed inclusion and exclusion criteria are provided in the Supplementary Methods. All healthy controls were randomly stratified, with 2,250 individuals allocated to the discovery set and 750 to the replication set, maintaining an approximate 3:1 ratio. Demographic and clinical characteristics of patients and controls are summarized in Supplementary Table 1 . Demographic and clinical assessment Demographic and clinical information was collected for all participants, including sex, age, age at onset, years of education, and current medication use. Symptom severity in patients was assessed with the Positive and Negative Syndrome Scale (PANSS). For analysis, we adopted the five-factor PANSS model, which provides a more nuanced characterization of symptom structure than the original three-subscale framework( 17 ). The five dimensions examined were positive, negative, disorganization, depression/anxiety, and hostility. Additional symptom domains were evaluated using the Hamilton Rating Scale for anxiety (HAMA), the Hamilton Rating Scale for Depression (HAMD), and the Young Mania Rating Scale (YMRS) for manic symptoms. Social functioning was further assessed with the Social Disability Screening Schedule (SDSS). Clustering analysis To classify patients into clinically homogeneous subtypes, we applied K-means cluster analysis to all 30 PANSS items in the discovery stage. Prior to clustering, item scores were standardized and adjusted for sex, age, age², and the sex-by-age interaction using linear regression to control the confound effect ( Supplementary Table 2 ). The optimal number of clusters (K) was determined using the NbClust R package, which selects the solution supported by the majority of 26 statistical indices( 18 ). The cluster stability was assessed with the Jaccard similarity coefficient ( \(\:{\gamma\:}_{c}\) ) via non-parametric bootstrap resampling (5,000 iterations, each using 80% of the sample). The mean Jaccard \(\:{\gamma\:}_{c}\) across iterations greater than 0.6 considered evidence of a robust clustering structure(19). To assess the stability and generalizability of the identified clusters, this procedure was performed independently in the replication cohort using the same parameter. Reproducibility of subtypes was evaluated by comparing symptom profiles across datasets. Within each cohort, Wilcoxon-Mann-Whitney (WMW) tests were performed for each PANSS item, comparing scores of individuals in a given subtype with all other patients. Spearman correlations were then calculated between WMW test estimates (z-score) for each subtype in the discovery dataset and its corresponding subtype in the replication dataset. Strong positive correlations (Spearman’s ρ > 0.7, P < 0.001) indicated successful replication and high similarity of clinical symptom patterns across cohorts. Comparison of symptom structures across the subtypes To compare symptom structures across the identified subtypes in the discovery dataset, we conducted network analysis. For each subtype, a network was estimated comprising nine nodes: the five PANSS factors, total scores from the HAMA, HAMD, and YMRS, and the mean SDSS score. The details for the analysis were displayed in Supplementary Methods . Genotyping and quality control Genomic DNA for all participants was extracted from peripheral blood and genotyped using the Infinium Asian Screening Array (ASA) chip ( https://support.illumina.com/downloads/infinium-asian-screening-array-v1-0-product-files.html ). For the replication cohort, genotype data were available for 400 patients. These samples, together with the entire discovery cohort, were subjected to quality control (QC) procedures. QC was conducted using PLINK v1.9 ( http://www.cog-genomics.org/plink ), and genotype imputation was performed with the ChinaMAP reference panel, a high-resolution panel specifically optimized for the Chinese population( 20 ). Details about QC and imputation procedures were provided in the Supplementary Methods .. After all QC steps, the final dataset comprised 2,746 patients (2,355 in the discovery cohort and 391 in the replication cohort) and 2,979 healthy controls (2,355 in the discovery cohort and 739 in the replication cohort). Genome-wide association analysis and meta-analysis GWAS were conducted for the full patient cohort as well as for each of the three clinical subtypes in both the discovery and replication datasets. Association testing was performed using logistic regression, adjusting for the first 10 PCs of ancestry to control for population stratification. Variants showing suggestive associations (P < 1 × 10⁻⁵) in the discovery stage were further evaluated in the replication stage, with P < 0.05 considered evidence of successful replication. Meta-analysis across the two stages was performed in PLINK v1.9 using the inverse-variance model. Manhattan and QQ plots were generated with the “CMplot” R package. Variants with a P value less than 1.25 × 10⁻⁸ (5 × 10⁻⁸ / 4) in the meta-analysis were considered statistically significant. Functional annotation To explore the potential functional implications of significant variants, we performed functional annotation using FUMA ( https://fuma.ctglab.nl/)(21) . Genomic risk loci were mapped to genes through both positional mapping and expression quantitative trait loci (eQTL) mapping. For eQTL mapping, we incorporated dorsolateral prefrontal cortex (DLPFC) data from PsychENCODE and whole blood data from GTEx to assess whether GWAS-identified variants were associated with gene expression levels ( 22 ). Gene-based analysis was conducted using Multi-marker Analysis of GenoMic Annotation (MAGMA)( 23 ), which tested associations across 18,601 protein-coding genes. The Bonferroni-corrected genome-wide significance threshold was set at P < 1.67 × 10⁻⁶ (0.05/18,601). To further investigate potential regulatory mechanisms of risk loci, we applied Hi-C coupled MAGMA (H-MAGMA)( 24 ), which integrates chromatin interaction profiles to map non-coding variants to their target genes. Six Hi-C datasets were analyzed, covering diverse brain-related cell types and developmental stages, including adult brain, fetal brain, cortical neurons, midbrain dopaminergic system, iPSC-derived neurons, and iPSC-derived astrocytes. Multiple testing correction was performed using the Bonferroni method. Genome-Wide and Pathway-Specific Polygenic Risk Score Calculation We calculated PRS using PRS-CSx, a cross-population polygenic prediction method based on a Bayesian regression framework( 25 ). This approach integrates GWAS summary statistics with linkage disequilibrium (LD) reference panels from multiple populations to improve predictive accuracy across ancestries. For this study, we incorporated ancestry-specific GWAS summary statistics from two populations (European and East Asian) provided by the Psychiatric Genomics Consortium (PGC). Full details of the data sources and analytic procedures are provided in the Supplementary Methods and Supplementary Table 3 . First, ancestry-specific posterior SNP effect sizes were estimated from GWAS summary statistics using PRS-CSx with default parameters. Based on these estimates, ancestry-specific genome-wide PRS (gPRS) were calculated for each individual using PLINK v1.9. A linear combination of ancestry-specific PRS that maximized predictive performance for case–control status in the validation dataset was then applied. The LD reference panels were derived from Phase 3 of the 1000 Genomes Project, with publicly available resources accessed at https://github.com/getian107/PRScsx . In addition to gPRS, we computed pathway-specific PRS (pPRS) and cell-type-specific PRS (ctPRS). These scores were generated by summing the effect sizes of variants located within gene sets ± 30 kb, a window chosen to capture most cis-regulatory variants and reduce potential mapping errors. Ancestry-specific GWAS summary statistics for SZ were used as reference data. For the pPRS analysis, we selected four neurotransmitter pathways that affect psychosis, including Glutamate, GABA, Dopamine, and Serotonin. For the ctPRS, we included gene sets of differentially expressed genes (DEGs) identified across six major brain cell types in the human prefrontal cortex from SZ versus healthy control comparisons( 26 ). These cell types included astrocytes (Ast), excitatory neurons (Ex), inhibitory neurons (In), microglia (Mic), oligodendrocytes (Oli), and oligodendrocyte progenitor cells (OPC). Details of the included gene sets are provided in the Supplementary Methods , with final gene lists in Supplementary Tables 4. For gPRS, pPRS, and ctPRS, logistic regression models were fitted to predict case–control status both across all patients and within each clinical subtype, adjusting for the first 10 ancestry PCs. To further explore subtype specificity, we also compared PRS between patients in each subtype and the remaining SZ patients. Multiple testing was controlled using Bonferroni correction, with associations considered statistically significant at corrected P < 0.05. Replication of previously published SNPs We further examined lead SNPs previously identified as genome-wide significant for SZ in East Asian (EAS) population (P < 5 × 10⁻⁸) ( 27 ). Of the 21 reported lead SNPs, 19 variants were available in our dataset after quality control. We evaluated the association of these variants both in the overall SZ cohort and within each clinically defined subtype. Results Clustering analysis results In the discovery stage, cluster analysis was conducted on 2,416 patients with SZ. Majority voting across 26 statistical indices suggested that either a two-cluster (supported by 7 indices) or three-cluster (supported by 7 indices) solution could be optimal. Both solutions demonstrated good stability (the \(\:{\gamma\:}_{c}\) in two-cluster model were 0.95 and 0.97, and \(\:{\gamma\:}_{c}\) in three-cluster model were 0.79, 0.79, and 0.85). The two-cluster solution primarily separated patients by overall symptom severity (low vs. high), whereas the three-cluster solution revealed an additional subgroup characterized by predominant negative and disorganized symptoms. Results are presented in Supplementary Fig. 1 , with detailed comparisons provided in Supplementary Results . Given that the three-cluster solution captured more clinically meaningful subgroups, offered greater interpretative value regarding symptom structure, and aligned with the recognized clinical subtypes of SZ( 28 , 29 ), we selected this model for subsequent analyses. The symptom profiles of the three clusters are shown in Fig. 1 A. Cluster-S (n = 306, 12.7%) was characterized by the most severe symptoms across nearly all domains, except for negative symptoms. Cluster-N (n = 711, 29.4%) was defined by predominant negative and disorganization symptoms. Cluster-L (n = 1,399, 57.9%) exhibited generally mild symptom severity. Clinical characteristics of the three subtypes are summarized in Supplementary Table 5 . Patients in Cluster-N had the highest mean age and the longest duration of illness compared to the other clusters. Both Cluster-S and Cluster-N were associated with poorer social functioning, as well as more severe manic, depressive, and anxiety symptoms, whereas patients in Cluster-L demonstrated comparatively preserved social functioning. To evaluate the reproducibility of the identified subtypes, we applied the same clustering procedure to the replication dataset (n = 480). This analysis again yielded three stable subtypes, with \(\:{\gamma\:}_{c}\) of 0.78, 0.79, and 0.76, indicating robust stability. The symptom structures and relative proportions of patients across the three subtypes were highly consistent with those observed in the discovery dataset (Fig. 1 B). Spearman correlation analysis of WMW estimates further confirmed strong concordance for Cluster-L and Cluster-N (Spearman’s ρ > 0.8), whereas concordance for Cluster-S was more moderate and did not reach the same high threshold (Fig. 1 C). Differences of symptom networks among the identified subtypes The network structures and centrality plots for the overall patient group and each subtype are shown in Supplementary Fig. 2 . Accuracy and stability analyses confirmed the robustness of all estimated networks ( Supplementary Figs. 3–4 ). In the full patient sample, the nodes with the highest strength centrality were negative symptoms, positive symptoms, and mania. However, the most central nodes differed across subtypes. In Cluster-S, depressive symptoms, anxiety, and mania emerged as the most central nodes; in Cluster-N, positive symptoms, depressive symptoms, and anxiety were most central; and in Cluster-L, mania, negative symptoms, and positive symptoms were most central. To formally test whether the three subtypes exhibited distinct symptom structures, we conducted network comparison tests. Significant differences were observed in overall network structure, node centrality, and edge weights across subtypes. Detailed results are provided in the Supplementary Results and Supplementary Table 7 . Genome-wide association studies of all schizophrenia and its clinical subtypes We conducted GWAS in the full SZ cohort and within each of the three subtypes, first in the discovery dataset and subsequently in the replication dataset. Principal component analysis confirmed that all participants were of East Asian ancestry ( Supplementary Fig. 5 ). In the discovery stage, 49 SNPs reached suggestive significance following LD-based clumping (20 in the full cohort, 6 in Cluster-S, 14 in Cluster-N, and 9 in Cluster-L). Of these, eight variants were successfully replicated (4 in the full cohort, 3 in Cluster-N, and 1 in Cluster-L). Meta-analyses integrating results across both stages identified four independent loci that achieved genome-wide significance (Table 1 ). The Manhattan and QQ plots for the meta-analyses are shown in Fig. 2 . For Cluster-S, the genomic inflation factor (λGC = 1.06) indicated only mild inflation, while LDSC analysis yielded an intercept of 1.05, suggesting that the inflation was attributable to polygenicity rather than confounding. Minimal inflation was observed in all other analyses (λGC ≤ 1.02). Table 1 Genome-Wide Association Results in Discovery, Replication, and Combined Samples. Cluster CHR Lead SNP A1 Nearest Gene Discovery stage Replication stage Meta-analysis OR P OR P OR P meta P HET All 21 rs2836330 G AP001422.3 1.55 2.39 × 10 − 10 1.46 1.58 × 10 − 03 1.53 1.64 × 10 − 12 0.640 All 3 rs9876206 C LRRC34 1.43 1.31 × 10 − 07 1.35 1.25 × 10 − 02 1.40 5.76 ×10 − 09 0.677 All 9 rs4240482 G DENND1A 1.38 3.42 × 10 − 06 1.30 3.37 × 10 − 02 1.36 3.61 × 10 − 07 0.667 All 21 rs220125 G UMODL1 0.75 9.87 × 10 − 06 0.77 2.14 × 10 − 02 0.75 6.37 × 10 − 07 0.863 N 1 rs3767295 C CNTN2 0.60 7.77 × 10 − 09 0.68 1.73 × 10 − 02 0.62 5.30 × 10 − 10 0.508 N 21 rs970115 C AP001422.3 1.53 1.43 × 10 − 06 1.70 8.21 × 10 − 04 1.57 5.29 × 10 − 09 0.548 N 1 rs56074668 C RP11-73M7.6:COL16A1 0.63 9.57 × 10 − 06 0.61 1.30 × 10 − 02 0.62 3.88 × 10 − 07 0.911 L 21 rs2836330 G AP001422.3 1.55 3.34 × 10 − 08 1.70 8.21 × 10 − 04 1.54 8.13E-10 0.843 Variant information and association statistics are shown for the most strongly associated SNP in each significant locus. Abbreviations: CHR: chromosome; SNP: single nucleotide polymorphism; A1: risk allele; OR: odds ratio; P: P values; P meta : P values for the meta-analysis of two-stage analysis; P HET : P values for Cochrane’s Q statistic In the meta-analysis of all patients, we identified two genome-wide significant loci ( Table 2 ). The first was located on chromosome 21 near the long non-coding RNA (lncRNA) gene AP001422.3 (lead SNP rs2836330, P = 1.64 × 10⁻¹², OR = 1.53). The second was on chromosome 3 within the intronic region of LRRC34 (lead SNP rs9876206, P = 5.76 × 10⁻⁹, OR = 1.40). Notably, the locus on chromosome 21 was also significant in Cluster-L (P = 8.59 × 10⁻¹¹, OR = 1.54). Furthermore, a nearby variant (rs970115) within the same gene ( AP001422.3 ) achieved genome-wide significance in Cluster-N (P = 5.29 × 10⁻⁹, OR = 1.57). In the subtype-specific meta-analyses, we identified one additional significant locus on chromosome 1, specific to Cluster-N. The lead SNP, rs3767295 (P = 5.30 × 10⁻¹⁰, OR = 0.62), was in an intronic region of the CNTN2 gene. This SNP showed only nominal significance (P < 0.05) in the other two subtypes and in the full patient analysis, with odds ratios closer to 1, indicating a markedly stronger effect in Cluster-N. Association results for all significant loci across subtypes are provided in Supplementary Fig. 6 , and the regional association plots are displayed in Supplementary Figs. 7–10 . Replication of previous EAS GWAS results We further examined the associations of previously reported SZ risk variants identified in EAS population ( 27 ). Of the 19 variants available in our dataset, 11 showed nominally significant associations (P < 0.05) in at least one analysis group. Importantly, the majority of these variants exhibited effect directions consistent with the original findings ( Supplementary Table 8 ). Within the combined patient group, 6 variants reached significance, of which only one was unique to this group, while the remaining 5 also showed associations in at least one subtype. Crucially, the subtype-based analyses revealed an additional 5 variants that were not significant in the overall patient group. These 5 variants were associated with only a single subtype, which 4 variants reached significance only in Cluster-S, and one only in Cluster-L. Gene-based analysis of all schizophrenia cases and its clinical subtypes Gene-based analyses using MAGMA and H-MAGMA identified four significant genes in the overall patient group and twelve significant genes in Cluster-N after Bonferroni correction (Fig. 3 ). In contrast, no genes reached statistical significance in Cluster-S or Cluster-L. In the overall patient group, three protein-coding genes ( UMODL1 , PRDM15 , and LRRC34 ) and one non-coding RNA gene ( C21orf128 ) reached genome-wide significance. In Cluster-N, eight protein-coding genes, including CNTN2 , DSTYK , LRRN2 , NFASC , RBBP5 , SLC2A13 , TMCC2 , and TMEM81 , and four RNA genes, including SNORD112 , TMCC2-AS1 , ENSG00000228153 , and ENSG00000240710 were identified. To further investigate potential regulatory mechanisms underlying these associations, we performed eQTL mapping analyses ( Supplementary Table 9 ). In whole blood, rs2836330 and rs970115 were identified as eQTLs for KCNJ15 . In the dorsolateral prefrontal cortex (DLPFC), rs3767295 was associated with CNTN2 expression (P FDR = 2.42 × 10⁻²). Additionally, rs9876206 was associated with the expression of four genes in the DLPFC, including MYNN (P FDR = 7.13 × 10⁻³), LRRIQ4 (P FDR = 3.76 × 10⁻²²), SAMD7 (P FDR = 5.23 × 10⁻³), and LRRC34 (P FDR = 2.30 × 10⁻¹²). Genome-wide PRS and pathway PRS analysis We first calculated gPRS for eight psychiatric disorders across all subtypes and healthy controls. As shown in Fig. 4 A, gPRS values for SCZ, BP were significantly elevated in all three subtypes relative to controls. Interestingly, the gPRS for MDD was significantly higher in Cluster-N (OR = 1.20 [1.07–1.35], P.adjust = 0.02), but not in Cluster-S (OR = 1.25 [1.07–1.46], P.adjust = 0.05) or Cluster-L (OR = 1.15 [1.04–1.28], P.adjust = 0.06). We next examined pPRS across four neurotransmitter pathways and ctPRS across six brain cell types, with results summarized in Fig. 4 B and 4 C. When comparing patients with controls, pPRS for glutamate were significantly elevated only in Cluster-L (OR = 1.19 [1.07–1.32], P.adjust = 0.011) and Cluster-N (OR = 1.27 [1.13–1.43], P.adjust = 0.001), but not in Cluster-S (OR = 1.22 [1.04–1.43], P.adjust = 0.156), and pPRS for GABA were only elevated in Cluster-N (OR = 1.19 [1.06–1.34], P.adjust = 0.032). For ctPRS, subtype-specific distinctions were also observed (Fig. 4 C). Astrocyte-specific PRS (Ast-PRS) was significantly elevated only in Cluster-S (OR = 1.29 [1.10–1.52], P.adjust = 0.035), whereas OPC-PRS was significantly elevated only in Cluster-N (OR = 1.25 [1.11–1.40], P.adjust = 0.005). Oligodendrocyte PRS (Oli-PRS) was elevated in both Cluster-S (OR = 1.31 [1.12–1.53], P.adjust = 0.012) and Cluster-N (OR = 1.19 [1.06–1.34], P.adjust = 0.045). These findings suggest distinct, cell-type-specific genetic risk architectures across the identified subtypes. Discussion This study identified three reproducible SZ subtypes based on PANSS symptom profiles. Subsequent genetic analyses, including subtype-specific GWAS, gene-based testing, and PRS analyses, revealed both unique risk loci and distinct genetic liability patterns across subtypes. Clinically, our analyses delineated three distinct subtypes: a general mild group (Cluster-L) with better overall functioning, a severe group (Cluster-S), and a group characterized by predominant negative and disorganized symptoms (Cluster-N). This three-cluster structure aligns closely with findings from previous studies that stratified SZ patients using PANSS scores, which reported comparable subtype in terms of both participant proportions and symptom profiles ( 12 , 28 , 29 ). These findings suggest that the three identified subtypes represent a robust and reproducible structure among SZ patients Symptom network analysis further highlighted structural differences among subtypes. Notably, we observed a dissociation between symptom severity and network centrality. For example, although Cluster-N displayed the most severe negative symptoms, the negative symptom nodes showed relatively low centrality within the network. This suggests that, for these patients, negative symptoms may operate as an isolated and entrenched core feature of illness rather than being dynamically interconnected with other symptom domains. This interpretation is consistent with prior studies of similar patient subgroups( 30 ) and may help explain the well-documented therapeutic resistance of negative symptoms in a subset of SZ patients( 31 ). Our analysis of the full SZ cohort identified two genome-wide significant loci: rs9876206 and rs2836330. The variant rs9876206 is located within LRRC34 , a gene encoding leucine-rich repeat containing, and play a role in DNA repair and telomere length regulation ( 32 ). However, the role of LRRC34 in SZ remains largely unexplored, underscoring the need for further mechanistic studies. The second significant variant, rs2836330, reached genome-wide significance in both the overall patient group and in Cluster-L. A secondary independent signal in the same region, rs970115, emerged as significant in Cluster-N. Both variants are eQTLs for KCNJ15 in whole blood. KCNJ15 encodes a potassium voltage-gated channel subunit and has previously been linked to Alzheimer’s disease( 33 ), epilepsy( 34 ), and Parkinson’s disease( 35 ). Elevated expression of KCNJ15 in white blood cells has also been reported in patients with atypical depression and psychotic symptoms( 36 ). However, its role in SZ has not been previously described, suggesting a novel avenue for investigation into its potential contribution to SZ pathophysiology. We also identified a genome-wide significant locus, rs3767295, located within CNTN2 , which was specific to Cluster-N. Notably, although previous large-scale GWAS have reported a suggestive association between SZ and another CNTN2 variant, rs11240341, which is in LD with rs3767295 in EAS population (R2 = 0.37) ( 2 , 37 , 38 ), no study to date has identified a CNTN2 variant that reaches genome-wide significance for SZ. This underscores the utility of patient stratification for enhancing statistical power and uncovering subtype-specific risk loci. CNTN2 encodes contactin-2, a neural cell adhesion molecule involved in axon guidance, myelination, and neural development( 39 ). Prior studies have reported reduced CNTN2 expression in the superior temporal gyrus of SZ patients( 40 ). Conversely, a proteome-wide Mendelian randomization study found that higher genetically predicted CNTN2 protein levels in cerebrospinal fluid were associated with reduced SZ risk( 41 ). Postmortem analyses further suggest that CNTN2 expression may vary by SZ subtype, showing increased expression in the amygdala of patients with disorganized SZ but decreased expression in those with paranoid SZ( 42 ). Our findings are consistent with this complexity. In Cluster-N, characterized by prominent negative and disorganized symptoms, the rs3767295-C allele was associated with both lower SZ risk and reduced CNTN2 expression in the dorsolateral prefrontal cortex. These results suggest that lower CNTN2 expression may be protective in certain subtypes, but could also represent compensatory mechanisms in others. Thus, the relationship between CNTN2 and SZ risk appears non-linear and context-dependent, varying across brain regions and clinical subtypes. When comparing gPRS across three severe psychiatric disorders, we observed a largely consistent pattern across the SZ subtypes, indicating a similar overall burden of common genetic risk. This is consistent with prior work showing that polygenic scores for SZ and BP did not differ significantly among psychosis biotypes( 43 ). Notably, however, a clear exception emerged for MDD-gPRS: only Cluster-N demonstrated a significantly elevated MDD-gPRS relative to controls. This finding is particularly compelling given the marked clinical overlap between negative symptoms, such as anhedonia and avolition, and the core features of depression( 44 , 45 ). Although a positive genetic correlation between SZ and MDD is well established( 46 ), previous studies have not reported an association between MDD polygenic liability and specific clinical subgroups within psychosis( 47 , 48 ). Our results suggest that Cluster-N may represent a subgroup of SZ patients with a distinct depression-related etiological pathway. This novel subtype-specific association highlights the potential of stratified approaches to reveal hidden genetic heterogeneity and warrants further investigation to clarify its biological mechanisms. In addition, we identified subtype-specific patterns in both pPRS and ctPRS. While certain pathways or cell types showed significantly elevated risk scores when all SZ patients were analyzed together, stratification into subtypes revealed distinct and more nuanced patterns. These differences were evident not only in comparisons between subtypes and healthy controls but also when contrasting individual subtypes against the remaining SZ patients. Such findings indicate that clinically defined subtypes may carry unique genetic liabilities tied to particular biological pathways and cellular processes. Unlike traditional gPRS, which aggregates genome-wide risk without functional context, pPRS provides a biologically informed framework by capturing genetic risk within specific gene sets or functional pathways. This improves interpretability and enhances its potential utility for disease stratification( 49 ). Indeed, prior work has demonstrated that pPRS outperforms gPRS in predicting endophenotypes and is associated with psychosis-related biotypes, underscoring its value for uncovering hidden biological heterogeneity in complex psychiatric disorders( 49 ). The present study has several notable strengths. We successfully identified and independently replicated clinically meaningful subtypes of SZ, providing strong evidence for their robustness. Moreover, by integrating subtype stratification with genetic analyses, we revealed distinct risk loci and biological patterns that were not detectable in traditional case–control analyses. Several limitations of this study should be acknowledged. The sample size, although substantial for a clinically detailed cohort, remains limited compared with recent large-scale case–control GWAS of SZ. This limitation reflects the inherent difficulty of assembling cohorts that combine genomic data with comprehensive symptom assessments. In addition, the cross-sectional design provides only a static snapshot of symptomatology and cannot capture longitudinal trajectories or changes in subtype membership over time. Finally, our analyses were restricted to clinical symptom dimensions, without incorporating other important domains such as cognitive function, neuroimaging, or biomarkers, which could provide further refinement of subtype definitions. Future work leveraging larger, multi-ancestry, deeply phenotyped datasets, ideally with longitudinal follow-up, will be critical to developing more comprehensive and stable subtypes of SZ. Conclusions Taken together, our findings demonstrate that stratifying schizophrenia patients into clinically homogeneous subtypes provides a powerful framework for uncovering subtype-specific genetic architectures that would otherwise remain obscured in conventional case–control analyses. The identification of novel loci, such as CNTN2 in the negative/disorganized subtype, alongside subtype-specific pathway and cell-type polygenic risks, underscores the biological validity of these subgroups. Importantly, this approach bridges clinical symptomatology with genetic mechanisms, offering a more nuanced understanding of the disorder’s heterogeneity. From a translational perspective, these insights raise the possibility that distinct subtypes may not only have different genetic underpinnings but could also differ in treatment response, disease trajectory, and prognosis. Future studies incorporating larger, multi-ancestry, and longitudinally phenotyped cohorts, as well as integrating multi-omics and neuroimaging data, will be crucial to refine these subtypes and link them to actionable biomarkers. Such efforts may pave the way toward precision psychiatry, where therapeutic strategies are tailored to biologically informed patient subgroups. Abbreviations Ast Astrocytes BD bipolar disorder ctPRS cell-type-specific polygenic risk scores eQTL expression quantitative trait loci Ex excitatory neurons gPRS genome-wide polygenic risk scores GWAS genome-wide association study HAMA Hamilton Rating Scale for anxiety HAMD Hamilton Rating Scale for Depression H-MAGMA Hi-C coupled MAGMA In inhibitory neurons LD linkage disequilibrium MAGMA Multi-marker Analysis of GenoMic Annotation MDD major depressive disorder Mic Microglia Oli oligodendrocytes OPC oligodendrocyte progenitor cells PANSS Positive and Negative Syndrome Scale PGC psychiatric genomics consortium pPRS pathway-specific polygenic risk scores PRS polygenic risk scores QC quality control SDSS Social Disability Screening Schedule SZ Schizophrenia WMW Wilcoxon-Mann-Whitney YMRS Young Mania Rating Scale Declarations Ethics approval and consent to participate This study was approved by the Ethic Committee of West China Hospital, Sichuan University (2018 − 185) and the Ethic Committee of the Affiliate Mental Health Center, Zhejiang University School of Medicine (2025-019). Consent for publication Not applicable Competing interests The authors declare no conflict of interest Funding: This work was supported by the China Brain Project (STI2030-2021ZD0200404 and STI2030-2021ZD0200800 to TL); the Key·R&D·by Hangzhou·Science and·Technology·Bureau (20241203A14 to T.L.); the Zhejiang Clinovation Pride (CXTD202501053 to T.L.). Author Contribution MZ, YZ, and TL conceived and designed the study; MZ, YZ, ZF, XW, HW, WW, HR, ML, QW, WD, and WG collected the data; MZ, YZ, and XD analyzed and interpretated the data; MZ made the figures, tables, and wrote the first draft; YZ, YZ, XL, and TL reviewed and revised. All authors read and approved the final paper. Acknowledgements We thank all participating patients and their families for taking part in the study. Data Availability The data underlying this article will be shared on reasonable request to the corresponding author. References Owen MJ, Legge SE. The nature of schizophrenia: As broad as it is long. Schizophr Res. 2022;242:109–12. Trubetskoy V, Pardiñas AF, Qi T, Panagiotaropoulou G, Awasthi S, Bigdeli TB, et al. Mapping genomic loci implicates genes and synaptic biology in schizophrenia. Nature. 2022;604(7906):502–8. Owen MJ, Legge SE, Rees E, Walters JTR, O'Donovan MC. Genomic findings in schizophrenia and their implications. Mol Psychiatry. 2023;28(9):3638–47. Taylor J, de Vries YA, van Loo HM, Kendler KS. Clinical characteristics indexing genetic differences in schizophrenia: a systematic review. Mol Psychiatry. 2023;28(2):883–90. Dworkin RH, Lenzenweger MF. Symptoms and the genetics of schizophrenia: implications for diagnosis. Am J Psychiatry. 1984;141(12):1541–6. Fanous AH, Zhou B, Aggen SH, Bergen SE, Amdur RL, Duan J, et al. Genome-wide association study of clinical dimensions of schizophrenia: polygenic effect on disorganized symptoms. Am J Psychiatry. 2012;169(12):1309–17. Pardiñas AF, Smart SE, Willcocks IR, Holmans PA, Dennison CA, Lynham AJ, et al. Interaction Testing and Polygenic Risk Scoring to Estimate the Association of Common Genetic Variants With Treatment Resistance in Schizophrenia. JAMA Psychiatry. 2022;79(3):260–9. Sada-Fuente E, Aranda S, Papiol S, Heilbronner U, Moltó MD, Aguilar EJ, et al. Common genetic variants contribute to heritability of age at onset of schizophrenia. Transl Psychiatry. 2023;13(1):201. Jablensky A. Subtyping schizophrenia: implications for genetic research. Mol Psychiatry. 2006;11(9):815–36. Warren TL, Tubbs JD, Lesh TA, Corona MB, Pakzad SS, Albuquerque MD, et al. Association of neurotransmitter pathway polygenic risk with specific symptom profiles in psychosis. Mol Psychiatry. 2024;29(8):2389–98. Holliday EG, McLean DE, Nyholt DR, Mowry BJ. Susceptibility locus on chromosome 1q23-25 for a schizophrenia subtype resembling deficit schizophrenia identified by latent class analysis. Arch Gen Psychiatry. 2009;66(10):1058–67. Dickinson D, Pratt DN, Giangrande EJ, Grunnagle M, Orel J, Weinberger DR, et al. Attacking Heterogeneity in Schizophrenia by Deriving Clinical Subgroups From Widely Available Symptom Data. Schizophr Bull. 2018;44(1):101–13. Dwyer DB, Kalman JL, Budde M, Kambeitz J, Ruef A, Antonucci LA, et al. An Investigation of Psychosis Subgroups With Prognostic Validation and Exploration of Genetic Underpinnings: The PsyCourse Study. JAMA Psychiatry. 2020;77(5):523–33. Mullins N, Forstner AJ, O'Connell KS, Coombes B, Coleman JRI, Qiao Z, et al. Genome-wide association study of more than 40,000 bipolar disorder cases provides new insights into the underlying biology. Nat Genet. 2021;53(6):817–29. Nguyen TD, Harder A, Xiong Y, Kowalec K, Hägg S, Cai N, et al. Genetic heterogeneity and subtypes of major depression. Mol Psychiatry. 2022;27(3):1667–75. Narita A, Nagai M, Mizuno S, Ogishima S, Tamiya G, Ueki M, et al. Clustering by phenotype and genome-wide association study in autism. Transl Psychiatry. 2020;10(1):290. Lim K, Peh OH, Yang Z, Rekhi G, Rapisarda A, See YM, et al. Large-scale evaluation of the Positive and Negative Syndrome Scale (PANSS) symptom architecture in schizophrenia. Asian J Psychiatr. 2021;62:102732. Charrad M, Ghazzali N, Boiteau V, Niknafs A, NbClust. An R Package for Determining the Relevant Number of Clusters in a Data Set. J Stat Softw. 2014;61(6):1–36. Hennig C. Cluster-wise assessment of cluster stability. Comput Stat Data Anal. 2007;52(1):258–71. Li L, Huang P, Sun X, Wang S, Xu M, Liu S, et al. The ChinaMAP reference panel for the accurate genotype imputation in Chinese populations. Cell Res. 2021;31(12):1308–10. Watanabe K, Taskesen E, van Bochoven A, Posthuma D. Functional mapping and annotation of genetic associations with FUMA. Nat Commun. 2017;8(1):1826. Wang D, Liu S, Warrell J, Won H, Shi X, Navarro FCP et al. Comprehensive functional genomic resource and integrative model for the human brain. Science. 2018;362(6420). de Leeuw CA, Mooij JM, Heskes T, Posthuma D. MAGMA: generalized gene-set analysis of GWAS data. PLoS Comput Biol. 2015;11(4):e1004219. Sey NYA, Hu B, Mah W, Fauni H, McAfee JC, Rajarajan P, et al. A computational tool (H-MAGMA) for improved prediction of brain-disorder risk genes by incorporating brain chromatin interaction profiles. Nat Neurosci. 2020;23(4):583–93. Ruan Y, Lin YF, Feng YA, Chen CY, Lam M, Guo Z, et al. Improving polygenic prediction in ancestrally diverse populations. Nat Genet. 2022;54(5):573–80. Ruzicka WB, Mohammadi S, Fullard JF, Davila-Velderrain J, Subburaju S, Tso DR, et al. Single-cell multi-cohort dissection of the schizophrenia transcriptome. Science. 2024;384(6698):eadg5136. Lam M, Chen CY, Li Z, Martin AR, Bryois J, Ma X, et al. Comparative genetic architectures of schizophrenia in East Asian and European populations. Nat Genet. 2019;51(12):1670–8. Lefort-Besnard J, Varoquaux G, Derntl B, Gruber O, Aleman A, Jardri R, et al. Patterns of schizophrenia symptoms: hidden structure in the PANSS questionnaire. Transl Psychiatry. 2018;8(1):237. Cowman M, Hodgekins J, Griffiths SL, Frawley E, O'Connor K, Fowler D et al. Cognitive and clinical profiles in first-episode psychosis and their relationship with functional outcomes. Br J Psychiatry. 2025:1–8. Demyttenaere K, Leenaerts N, Acsai K, Sebe B, Laszlovszky I, Barabássy Á, et al. Disentangling the symptoms of schizophrenia: Network analysis in acute phase patients and in patients with predominant negative symptoms. Eur Psychiatry. 2021;65(1):e18. Leucht S, Chaimani A, Krause M, Schneider-Thoma J, Wang D, Dong S, et al. The response of subgroups of patients with schizophrenia to different antipsychotic drugs: a systematic review and meta-analysis. Lancet Psychiatry. 2022;9(11):884–93. Pathare ADS, Džigurski J, Pujol-Gualdo N, Rukins V, Peters M, Mägi R, et al. A large-scale genome-wide association study on female genital tract polyps highlights role of DNA repair, cell proliferation, and cell growth. Hum Reprod. 2025;40(4):750–63. Zhou X, Chen Y, Mok KY, Zhao Q, Chen K, Chen Y, et al. Identification of genetic risk factors in the Chinese population implicates a role of immune system in Alzheimer's disease pathogenesis. Proc Natl Acad Sci U S A. 2018;115(8):1697–706. Wang S, Li Z, Ding X, Zhao Z, Zhang M, Xu H, et al. Integrative Analyses Identify KCNJ15 as a Candidate Gene in Patients with Epilepsy. Neurol Ther. 2022;11(4):1767–76. Chen X, Finol-Urdaneta RK, Chen M, Sykes AM, Gao B, Iqbal J, et al. Parkinson's disease-linked Kir4.2 mutation R28C leads to loss of ion channel function. J Physiol. 2025;603(12):3499–518. Ahn I, Chang S, Choi S-H, Han J, Kim Y. Exploration of Novel Biomarkers through a Precision Medicine Approach Using Multi-omics and Brain Organoids in Patients with Atypical Depression and Psychotic Symptoms. bioRxiv. 2025:2025.04.21.649748. Li ZA-O, Chen J, Yu H, He L, Xu Y, Zhang D, et al. Genome-wide association analysis identifies 30 new susceptibility loci for schizophrenia. Nat Genet. 2017;49(11):1576–83. Ikeda M, Takahashi A, Kamatani Y, Momozawa Y, Saito T, Kondo K, et al. Genome-Wide Association Study Detected Novel Susceptibility Genes for Schizophrenia and Shared Trans-Populations/Diseases Genetic Effect. Schizophr Bull. 2019;45(4):824–34. Fan S, Liu J, Chofflet N, Bailey AO, Russell WK, Zhang Z, et al. Molecular mechanism of contactin 2 homophilic interaction. Structure. 2024;32(10):1652–66. Roussos P, Katsel P, Fau - Davis KL, Davis Kl Fau -, Bitsios P, Bitsios P, Fau - Giakoumaki SG, Giakoumaki Sg Fau - Jogia J, Jogia J, Fau - Rozsnyai K et al. Molecular and genetic evidence for abnormalities in the nodes of Ranvier in schizophrenia. Arch Gen Psychiatry. 2012;69(1):7–15. Gu X, Dou M, Su W, Jiang Z, Duan Q, Cao B, et al. Identifying novel proteins underlying schizophrenia via integrating pQTLs of the plasma, CSF, and brain with GWAS summary data. BMC Med. 2022;20(1):474. Chang X, Liu Y, Hahn CG, Gur RE, Sleiman PMA, Hakonarson H. RNA-seq analysis of amygdala tissue reveals characteristic expression profiles in schizophrenia. Transl Psychiatry. 2017;7(8):e1203. Xia CA-O, Alliey-Rodriguez NA-O, Tamminga CA-O, Keshavan MA-OX, Pearlson GD, Keedy SK, et al. Genetic analysis of psychosis Biotypes: shared Ancestry-adjusted polygenic risk and unique genomic associations. Mol Psychiatry. 2025;30(6):2673–85. Guessoum SB, Le Strat Y, Dubertret C, Mallet J. A transnosographic approach of negative symptoms pathophysiology in schizophrenia and depressive disorders. Prog Neuropsychopharmacol Biol Psychiatry. 2020;99(Electronic):1878–4216. Krynicki CA-O, Upthegrove RA-O, Deakin JFW, Barnes TA-OX. The relationship between negative symptoms and depression in schizophrenia: a systematic review. Acta Psychiatr Scand. 2018;137(5):380–90. Romero CA-O, Werme JA-O, Jansen PA-O, Gelernter JA-O, Stein MA-O, Levey DA-O, et al. Exploring the genetic overlap between twelve psychiatric disorders. Nat Genet. 2022;54(12):1795–802. Legge SE, Cardno AG, Allardyce J, Dennison C, Hubbard L, Pardiñas AF, et al. Associations Between Schizophrenia Polygenic Liability, Symptom Dimensions, and Cognitive Ability in Schizophrenia. JAMA Psychiatry. 2021;78(10):1143–51. Ahangari MA-O, Bustamante D, Kirkpatrick R, Nguyen TH, Verrelli BC, Fanous A et al. Relationship between polygenic risk scores and symptom dimensions of schizophrenia and schizotypy in multiplex families with schizophrenia. Br J Psychiatry.223(1):301–8. Choi SW, García-González JA-OX, Ruan YA-O, Wu HM, Porras CA-O, Johnson JA-O, et al. PRSet: Pathway-based polygenic risk score analyses and software. PLoS Genet. 2023;19(2):e1010624. Additional Declarations No competing interests reported. Supplementary Files SupplementaryMaterial.docx Cite Share Download PDF Status: Under Review Version 1 posted Editorial decision: Revision requested 27 Feb, 2026 Reviews received at journal 23 Feb, 2026 Reviews received at journal 18 Feb, 2026 Reviewers agreed at journal 04 Feb, 2026 Reviewers agreed at journal 04 Feb, 2026 Reviewers invited by journal 02 Feb, 2026 Editor invited by journal 28 Jan, 2026 Editor assigned by journal 19 Jan, 2026 Submission checks completed at journal 19 Jan, 2026 First submitted to journal 16 Jan, 2026 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-8620966","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":585196199,"identity":"be68c06c-1afd-43c2-bbe7-3d89335bc1ca","order_by":0,"name":"Meng Zhou","email":"","orcid":"","institution":"Affiliated Mental Health Center \u0026 Hangzhou Seventh People’s Hospital, Zhejiang University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Meng","middleName":"","lastName":"Zhou","suffix":""},{"id":585196200,"identity":"e2a939d9-55e1-43cc-b214-e303edbaf872","order_by":1,"name":"Yamin Zhang","email":"","orcid":"","institution":"Affiliated Mental Health Center \u0026 Hangzhou Seventh People’s Hospital, Zhejiang University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Yamin","middleName":"","lastName":"Zhang","suffix":""},{"id":585196201,"identity":"43bd1f08-3ae1-457d-9065-9b1b6d03681c","order_by":2,"name":"Yilu Zhao","email":"","orcid":"","institution":"Affiliated Mental Health Center \u0026 Hangzhou Seventh People’s Hospital, Zhejiang University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Yilu","middleName":"","lastName":"Zhao","suffix":""},{"id":585196202,"identity":"698411fa-069c-4171-bf33-42fb4ab64001","order_by":3,"name":"Zheng Fu","email":"","orcid":"","institution":"School of Mental Health, Wenzhou Medical University","correspondingAuthor":false,"prefix":"","firstName":"Zheng","middleName":"","lastName":"Fu","suffix":""},{"id":585196204,"identity":"44b85fe3-4664-47bb-9854-24ea5c1bb3cf","order_by":4,"name":"Xiaohui Wang","email":"","orcid":"","institution":"Affiliated Mental Health Center \u0026 Hangzhou Seventh People’s Hospital, Zhejiang University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Xiaohui","middleName":"","lastName":"Wang","suffix":""},{"id":585196205,"identity":"e27b195b-92c2-4d8c-a9c9-a58aba57f971","order_by":5,"name":"Xinglun Dang","email":"","orcid":"","institution":"Zhongda Hospital, School of Medicine, Advanced Institute for Life and Health, Jiangsu Provincial Key Laboratory of Brain Science and Medicine, Southeast University","correspondingAuthor":false,"prefix":"","firstName":"Xinglun","middleName":"","lastName":"Dang","suffix":""},{"id":585196206,"identity":"729c1190-1efb-41eb-9f49-afbc240268ef","order_by":6,"name":"Huiyao Wang","email":"","orcid":"","institution":"Mental Health Center and Psychiatric Laboratory, The State Key Laboratory of Biotherapy, West China Hospital of Sichuan University","correspondingAuthor":false,"prefix":"","firstName":"Huiyao","middleName":"","lastName":"Wang","suffix":""},{"id":585196207,"identity":"89799330-14f2-404a-93bc-1b368fef8fb5","order_by":7,"name":"Wei Wei","email":"","orcid":"","institution":"Affiliated Mental Health Center \u0026 Hangzhou Seventh People’s Hospital, Zhejiang University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Wei","middleName":"","lastName":"Wei","suffix":""},{"id":585196208,"identity":"d5dcfe20-68ac-4fb0-9fbd-69c9d24b860c","order_by":8,"name":"Hongyan Ren","email":"","orcid":"","institution":"Mental Health Center and Psychiatric Laboratory, The State Key Laboratory of Biotherapy, West China Hospital of Sichuan University","correspondingAuthor":false,"prefix":"","firstName":"Hongyan","middleName":"","lastName":"Ren","suffix":""},{"id":585196209,"identity":"51dd2a7f-f5dc-44f5-9332-49384caea79c","order_by":9,"name":"Mingli Li","email":"","orcid":"","institution":"Mental Health Center and Psychiatric Laboratory, The State Key Laboratory of Biotherapy, West China Hospital of Sichuan University","correspondingAuthor":false,"prefix":"","firstName":"Mingli","middleName":"","lastName":"Li","suffix":""},{"id":585196210,"identity":"dae44128-f174-43ed-a3e2-15daa06ca7db","order_by":10,"name":"Qiang Wang","email":"","orcid":"","institution":"Mental Health Center and Psychiatric Laboratory, The State Key Laboratory of Biotherapy, West China Hospital of Sichuan University","correspondingAuthor":false,"prefix":"","firstName":"Qiang","middleName":"","lastName":"Wang","suffix":""},{"id":585196211,"identity":"a83a1397-f18f-431a-ba58-3c80d77a3e75","order_by":11,"name":"Wei Deng","email":"","orcid":"","institution":"Affiliated Mental Health Center \u0026 Hangzhou Seventh People’s Hospital, Zhejiang University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Wei","middleName":"","lastName":"Deng","suffix":""},{"id":585196212,"identity":"133d738e-4645-4ad9-8835-ce139864d938","order_by":12,"name":"Wanjun Guo","email":"","orcid":"","institution":"Affiliated Mental Health Center \u0026 Hangzhou Seventh People’s Hospital, Zhejiang University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Wanjun","middleName":"","lastName":"Guo","suffix":""},{"id":585196213,"identity":"b99346fc-d123-4482-bc9a-5f621557f821","order_by":13,"name":"Xiongjian Luo","email":"","orcid":"","institution":"Zhongda Hospital, School of Medicine, Advanced Institute for Life and Health, Jiangsu Provincial Key Laboratory of Brain Science and Medicine, Southeast University","correspondingAuthor":false,"prefix":"","firstName":"Xiongjian","middleName":"","lastName":"Luo","suffix":""},{"id":585196214,"identity":"cce38a6f-ba24-40e7-b600-26c57683fc00","order_by":14,"name":"Tao Li","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA7ElEQVRIiWNgGAWjYBACPmYwxQbEzAcYJEDsAwS0sDEzw7SwJRCphYEZxuQxgNAEtbDzH/zw4Q9f4vz2M58/WLYxyPHdSGD8XIDfYcySM3jYEjecyd1gINnGYCx5I4FZegYBv0jzSAC1SPBuSABqSdxwI4GNmYeALb95DNgS58/geXAAqKWeGC1s0jwJbIkNN3gYG4BaEgyI0GJmOeMAm/GGM2nGDBLnJAxnnnnYLI1PCz//wcc3Pvw5Jju//fDjzxJlNvJ8x5MPfsanBQqOgUlmCXBkMjYQ1sDAUAMmGT8Qo3YUjIJRMApGHAAAIsVBMdgIOPwAAAAASUVORK5CYII=","orcid":"","institution":"Affiliated Mental Health Center \u0026 Hangzhou Seventh People’s Hospital, Zhejiang University School of Medicine","correspondingAuthor":true,"prefix":"","firstName":"Tao","middleName":"","lastName":"Li","suffix":""}],"badges":[],"createdAt":"2026-01-16 16:38:37","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8620966/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8620966/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":101940384,"identity":"9aa8e149-03f3-41b0-9cb2-6a1aefb48f9b","added_by":"auto","created_at":"2026-02-05 09:14:03","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":593142,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePatient Subtyping and Symptom Network Analysis in Schizophrenia. (A) \u003c/strong\u003eA 3D scatter plot showing three patient clusters (Cluster-L, Cluster-N, Cluster-S) based on scores form the three subscales of PANSS. \u003cstrong\u003e(B)\u003c/strong\u003e A radar plot comparing the z-scores of five symptom dimensions across the three subtypes. (C) Spearman correlation of WMW estimates of every subtype in the discovery dataset and replication dataset.\u003c/p\u003e","description":"","filename":"image1.png","url":"https://assets-eu.researchsquare.com/files/rs-8620966/v1/9a9522fe1960a55011771a4b.png"},{"id":101940319,"identity":"bfbb2fcf-df78-46d1-a5c9-c64ea6ef25d6","added_by":"auto","created_at":"2026-02-05 09:13:42","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":932998,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eGenome-Wide Association Study Results.\u003c/strong\u003e\u0026nbsp; Manhattan plot (left) and Quantile-Quantile (Q-Q) plot (right) for (A) All patients, (B) Cluster-S, (C) Cluster-N, and (D) Cluster-L. X-axis shows chromosomal positions. In the Manhattan plot, the Y-axis shows –log10 p values. The dash line on the top indicates the genome-wide significance threshold (p=5.0*10\u003csup\u003e-8\u003c/sup\u003e), and the dash line on the bottom indicates the suggestive genome-wide significance threshold (p=1.0*10\u003csup\u003e-5\u003c/sup\u003e). The Q-Q plot indicated the observed versus the expected -log10 transformed P-values under the null hypothesis.\u003c/p\u003e","description":"","filename":"image2.png","url":"https://assets-eu.researchsquare.com/files/rs-8620966/v1/2195a5bce8d39bea80300ae0.png"},{"id":101940380,"identity":"be090fa7-7230-40d1-9642-ad04cdf3db15","added_by":"auto","created_at":"2026-02-05 09:13:59","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":183772,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eGene-Based Analysis Results Across Different Subtypes.\u003c/strong\u003e This bubble plot represents the results of gene-based analysis using MAGMA and H-MAGMA with different brain-related dataset. The Y-axis represents genes, while the X-axis represents different datasets. Color intensity corresponds to the -log\u003csub\u003e10 \u003c/sub\u003e(P\u003csub\u003eadjust\u003c/sub\u003e) value, with larger bubbles indicating larger Z statistics. Results are shown separately for all patients and Cluster-N. *: P\u003csub\u003eadjust\u003c/sub\u003e \u0026lt; 0.05.\u003c/p\u003e","description":"","filename":"image3.png","url":"https://assets-eu.researchsquare.com/files/rs-8620966/v1/0859ded8a2206d4cb9b4b7d4.png"},{"id":101940366,"identity":"84f0d1e5-0db5-443d-9c2b-ccbc74949b6f","added_by":"auto","created_at":"2026-02-05 09:13:55","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":184381,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePolygenic Risk Score (PRS) Effect Sizes for Subtypes. \u003c/strong\u003eForest plots showing the odds ratio (OR) for the association between various PRS and patient groups (Cluster-L, Cluster-S, Cluster-N) relative to healthy controls. \u003cstrong\u003e(A)\u003c/strong\u003e Genome-wide PRS for three psychiatric disorders. \u003cstrong\u003e(B) \u003c/strong\u003ePathway-based PRS. \u003cstrong\u003e(C)\u003c/strong\u003e cell-type-specific PRS. Data are presented as OR with 95% confidence intervals. Solid circles indicate that Bonferroni-P \u0026lt; 0.05.\u003c/p\u003e\n\u003cp\u003eSCZ: Schizophrenia; BP: Bipolar Disorder; MDD: Major Depressive Disorder; Ast: Astrocyte; Ex: Excitatory Neurons; In: Inhibitory Neurons; Mic: Microglia; Oli: Oligodendrocytes; OPC: Oligodendrocyte Progenitor Cells\u003c/p\u003e","description":"","filename":"image4.png","url":"https://assets-eu.researchsquare.com/files/rs-8620966/v1/66bcd3bfb5a75080049eb9ae.png"},{"id":101943724,"identity":"60b00644-978a-404a-af0e-ab88539516b4","added_by":"auto","created_at":"2026-02-05 09:43:00","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2899570,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8620966/v1/34127df0-fbd4-4d4e-8dd8-e180963dd1ab.pdf"},{"id":101940371,"identity":"aebee45b-1f46-47b2-8804-9ce0f213ebea","added_by":"auto","created_at":"2026-02-05 09:13:56","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":4291203,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryMaterial.docx","url":"https://assets-eu.researchsquare.com/files/rs-8620966/v1/5085d02b6b07e15bb7963155.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Genetic Architecture of Schizophrenia Clinical Subtypes","fulltext":[{"header":"Background","content":"\u003cp\u003eSchizophrenia (SZ) is a highly heritable psychotic disorder characterized by both complex genetic architecture and diverse clinical manifestations(\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e). Large-scale genome-wide association studies (GWAS) have identified more than 200 risk loci associated with SZ(\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e). However, these findings explain only approximately 40% of the estimated heritability, leaving a substantial proportion unaccounted for(\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eA major barrier to resolving this \u0026ldquo;missing heritability\u0026rdquo; is the marked heterogeneity in both clinical presentation and genetic architecture. Distinct symptom profiles and treatment responses may reflect different genetic liabilities(\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e). For example, severe negative and disorganized symptoms have been linked to stronger family histories of psychiatric illness(\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e) and higher SZ polygenic risk scores (PRS)(\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e). Elevated PRS have also been associated with treatment resistance(\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e) and earlier onset(\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e). These findings underscore the limitations of treating SZ as a single, uniform entity in genetic analyses.\u003c/p\u003e \u003cp\u003eSubtyping patients into more clinically homogeneous groups offers a promising avenue for genetic discovery(\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e). Several studies have stratified patients based on symptom dimensions, cognitive impairments, or treatment response, identifying subgroups with distinct genetic profiles(\u003cspan additionalcitationids=\"CR11\" citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e). For instance, variants in the chromosome 1q region showed significant associations with the negative symptom subtype but not when all patients were analyzed together(\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e). Beyond SZ, phenotype-driven stratification has also revealed distinct liabilities across the broader psychosis spectrum(\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e). Collectively, these findings suggest that identifying clinically meaningful subgroups could help bridge the gap in unexplained heritability, clarify biological mechanisms, and ultimately inform more precise therapeutic strategies for SZ.\u003c/p\u003e \u003cp\u003eDespite increasing interest in the genetic underpinnings of SZ subtypes, most existing research have focused on genome-wide PRS (gPRS) to estimate overall liability. While informative, such approaches provide limited insight into the specific biological processes that differentiate subtypes. In contrast, subtype-specific GWAS have proven powerful in other psychiatric disorders, including bipolar disorder(\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e), major depressive disorder(\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e), and autism(\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e), where they have revealed subgroup-specific genetic architectures that would otherwise be masked in aggregated analyses. In SZ, however, subtype-based GWAS remain relatively rare. Expanding the application of subtype-focused genetic analyses in SZ is therefore crucial for uncovering novel risk loci and clarifying disease mechanisms.\u003c/p\u003e \u003cp\u003eIn this study, we sought to address this gap by investigating the genetic heterogeneity underlying clinically defined subtypes of SZ. We first classified patients into distinct clinical subtypes based on their symptom profiles and examined their symptom network structures. We then conducted subtype-specific GWAS and gene-based analyses to identify loci and genes associated with particular subgroups. Finally, we evaluated whether genetic risks, both genome-wide and within specific biological pathways and brain cell types, varied across subtypes. Through this approach, we aimed to demonstrate that stratification by clinical subtype can enhance the discovery of biologically meaningful genetic signals that are obscured in traditional case\u0026ndash;control GWAS.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eStudy subjects\u003c/h2\u003e \u003cp\u003eA total of 2,896 patients with SZ were recruited from eight psychiatric hospitals across Sichuan and Zhejiang provinces in China, along with 3,000 healthy controls recruited from the community. Recruitment occurred in two phases. The discovery cohort comprised 2,410 patients enrolled between 2021 and 2022, while an additional 480 patients were recruited between 2023 and 2025 for replication. All participants were of Han Chinese ancestry, and detailed inclusion and exclusion criteria are provided in the \u003cb\u003eSupplementary Methods.\u003c/b\u003e All healthy controls were randomly stratified, with 2,250 individuals allocated to the discovery set and 750 to the replication set, maintaining an approximate 3:1 ratio. Demographic and clinical characteristics of patients and controls are summarized in \u003cb\u003eSupplementary Table\u0026nbsp;1\u003c/b\u003e.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eDemographic and clinical assessment\u003c/h3\u003e\n\u003cp\u003eDemographic and clinical information was collected for all participants, including sex, age, age at onset, years of education, and current medication use. Symptom severity in patients was assessed with the Positive and Negative Syndrome Scale (PANSS). For analysis, we adopted the five-factor PANSS model, which provides a more nuanced characterization of symptom structure than the original three-subscale framework(\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e). The five dimensions examined were positive, negative, disorganization, depression/anxiety, and hostility. Additional symptom domains were evaluated using the Hamilton Rating Scale for anxiety (HAMA), the Hamilton Rating Scale for Depression (HAMD), and the Young Mania Rating Scale (YMRS) for manic symptoms. Social functioning was further assessed with the Social Disability Screening Schedule (SDSS).\u003c/p\u003e\n\u003ch3\u003eClustering analysis\u003c/h3\u003e\n\u003cp\u003eTo classify patients into clinically homogeneous subtypes, we applied K-means cluster analysis to all 30 PANSS items in the discovery stage. Prior to clustering, item scores were standardized and adjusted for sex, age, age\u0026sup2;, and the sex-by-age interaction using linear regression to control the confound effect (\u003cb\u003eSupplementary Table\u0026nbsp;2\u003c/b\u003e). The optimal number of clusters (K) was determined using the \u003cem\u003eNbClust\u003c/em\u003e R package, which selects the solution supported by the majority of 26 statistical indices(\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e). The cluster stability was assessed with the Jaccard similarity coefficient (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\gamma\\:}_{c}\\)\u003c/span\u003e\u003c/span\u003e) via non-parametric bootstrap resampling (5,000 iterations, each using 80% of the sample). The mean Jaccard \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\gamma\\:}_{c}\\)\u003c/span\u003e\u003c/span\u003e across iterations greater than 0.6 considered evidence of a robust clustering structure(19). To assess the stability and generalizability of the identified clusters, this procedure was performed independently in the replication cohort using the same parameter. Reproducibility of subtypes was evaluated by comparing symptom profiles across datasets. Within each cohort, Wilcoxon-Mann-Whitney (WMW) tests were performed for each PANSS item, comparing scores of individuals in a given subtype with all other patients. Spearman correlations were then calculated between WMW test estimates (z-score) for each subtype in the discovery dataset and its corresponding subtype in the replication dataset. Strong positive correlations (Spearman\u0026rsquo;s ρ\u0026thinsp;\u0026gt;\u0026thinsp;0.7, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001) indicated successful replication and high similarity of clinical symptom patterns across cohorts.\u003c/p\u003e\n\u003ch3\u003eComparison of symptom structures across the subtypes\u003c/h3\u003e\n\u003cp\u003eTo compare symptom structures across the identified subtypes in the discovery dataset, we conducted network analysis. For each subtype, a network was estimated comprising nine nodes: the five PANSS factors, total scores from the HAMA, HAMD, and YMRS, and the mean SDSS score. The details for the analysis were displayed in \u003cb\u003eSupplementary Methods\u003c/b\u003e.\u003c/p\u003e\n\u003ch3\u003eGenotyping and quality control\u003c/h3\u003e\n\u003cp\u003eGenomic DNA for all participants was extracted from peripheral blood and genotyped using the Infinium Asian Screening Array (ASA) chip (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://support.illumina.com/downloads/infinium-asian-screening-array-v1-0-product-files.html\u003c/span\u003e\u003cspan address=\"https://support.illumina.com/downloads/infinium-asian-screening-array-v1-0-product-files.html\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). For the replication cohort, genotype data were available for 400 patients. These samples, together with the entire discovery cohort, were subjected to quality control (QC) procedures.\u003c/p\u003e \u003cp\u003eQC was conducted using PLINK v1.9 (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://www.cog-genomics.org/plink\u003c/span\u003e\u003cspan address=\"http://www.cog-genomics.org/plink\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e), and genotype imputation was performed with the ChinaMAP reference panel, a high-resolution panel specifically optimized for the Chinese population(\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e). Details about QC and imputation procedures were provided in the \u003cb\u003eSupplementary Methods\u003c/b\u003e.. After all QC steps, the final dataset comprised 2,746 patients (2,355 in the discovery cohort and 391 in the replication cohort) and 2,979 healthy controls (2,355 in the discovery cohort and 739 in the replication cohort).\u003c/p\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eGenome-wide association analysis and meta-analysis\u003c/h2\u003e \u003cp\u003eGWAS were conducted for the full patient cohort as well as for each of the three clinical subtypes in both the discovery and replication datasets. Association testing was performed using logistic regression, adjusting for the first 10 PCs of ancestry to control for population stratification. Variants showing suggestive associations (P\u0026thinsp;\u0026lt;\u0026thinsp;1 \u0026times; 10⁻⁵) in the discovery stage were further evaluated in the replication stage, with P\u0026thinsp;\u0026lt;\u0026thinsp;0.05 considered evidence of successful replication. Meta-analysis across the two stages was performed in PLINK v1.9 using the inverse-variance model. Manhattan and QQ plots were generated with the \u0026ldquo;CMplot\u0026rdquo; R package. Variants with a P value less than 1.25 \u0026times; 10⁻⁸ (5 \u0026times; 10⁻⁸ / 4) in the meta-analysis were considered statistically significant.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eFunctional annotation\u003c/h3\u003e\n\u003cp\u003eTo explore the potential functional implications of significant variants, we performed functional annotation using FUMA (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://fuma.ctglab.nl/)(21)\u003c/span\u003e\u003cspan address=\"https://fuma.ctglab.nl/)(21)\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Genomic risk loci were mapped to genes through both positional mapping and expression quantitative trait loci (eQTL) mapping. For eQTL mapping, we incorporated dorsolateral prefrontal cortex (DLPFC) data from PsychENCODE and whole blood data from GTEx to assess whether GWAS-identified variants were associated with gene expression levels (\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eGene-based analysis was conducted using Multi-marker Analysis of GenoMic Annotation (MAGMA)(\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e), which tested associations across 18,601 protein-coding genes. The Bonferroni-corrected genome-wide significance threshold was set at P\u0026thinsp;\u0026lt;\u0026thinsp;1.67 \u0026times; 10⁻⁶ (0.05/18,601). To further investigate potential regulatory mechanisms of risk loci, we applied Hi-C coupled MAGMA (H-MAGMA)(\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e), which integrates chromatin interaction profiles to map non-coding variants to their target genes. Six Hi-C datasets were analyzed, covering diverse brain-related cell types and developmental stages, including adult brain, fetal brain, cortical neurons, midbrain dopaminergic system, iPSC-derived neurons, and iPSC-derived astrocytes. Multiple testing correction was performed using the Bonferroni method.\u003c/p\u003e\n\u003ch3\u003eGenome-Wide and Pathway-Specific Polygenic Risk Score Calculation\u003c/h3\u003e\n\u003cp\u003eWe calculated PRS using PRS-CSx, a cross-population polygenic prediction method based on a Bayesian regression framework(\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e). This approach integrates GWAS summary statistics with linkage disequilibrium (LD) reference panels from multiple populations to improve predictive accuracy across ancestries. For this study, we incorporated ancestry-specific GWAS summary statistics from two populations (European and East Asian) provided by the Psychiatric Genomics Consortium (PGC). Full details of the data sources and analytic procedures are provided in the \u003cb\u003eSupplementary Methods\u003c/b\u003e and \u003cb\u003eSupplementary Table\u0026nbsp;3\u003c/b\u003e.\u003c/p\u003e \u003cp\u003eFirst, ancestry-specific posterior SNP effect sizes were estimated from GWAS summary statistics using PRS-CSx with default parameters. Based on these estimates, ancestry-specific genome-wide PRS (gPRS) were calculated for each individual using PLINK v1.9. A linear combination of ancestry-specific PRS that maximized predictive performance for case\u0026ndash;control status in the validation dataset was then applied. The LD reference panels were derived from Phase 3 of the 1000 Genomes Project, with publicly available resources accessed at \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://github.com/getian107/PRScsx\u003c/span\u003e\u003cspan address=\"https://github.com/getian107/PRScsx\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/p\u003e \u003cp\u003eIn addition to gPRS, we computed pathway-specific PRS (pPRS) and cell-type-specific PRS (ctPRS). These scores were generated by summing the effect sizes of variants located within gene sets\u0026thinsp;\u0026plusmn;\u0026thinsp;30 kb, a window chosen to capture most cis-regulatory variants and reduce potential mapping errors. Ancestry-specific GWAS summary statistics for SZ were used as reference data.\u003c/p\u003e \u003cp\u003eFor the pPRS analysis, we selected four neurotransmitter pathways that affect psychosis, including Glutamate, GABA, Dopamine, and Serotonin. For the ctPRS, we included gene sets of differentially expressed genes (DEGs) identified across six major brain cell types in the human prefrontal cortex from SZ versus healthy control comparisons(\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e). These cell types included astrocytes (Ast), excitatory neurons (Ex), inhibitory neurons (In), microglia (Mic), oligodendrocytes (Oli), and oligodendrocyte progenitor cells (OPC). Details of the included gene sets are provided in the \u003cb\u003eSupplementary Methods\u003c/b\u003e, with final gene lists in \u003cb\u003eSupplementary Tables\u0026nbsp;4.\u003c/b\u003e\u003c/p\u003e \u003cp\u003eFor gPRS, pPRS, and ctPRS, logistic regression models were fitted to predict case\u0026ndash;control status both across all patients and within each clinical subtype, adjusting for the first 10 ancestry PCs. To further explore subtype specificity, we also compared PRS between patients in each subtype and the remaining SZ patients. Multiple testing was controlled using Bonferroni correction, with associations considered statistically significant at corrected P\u0026thinsp;\u0026lt;\u0026thinsp;0.05.\u003c/p\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eReplication of previously published SNPs\u003c/h2\u003e \u003cp\u003eWe further examined lead SNPs previously identified as genome-wide significant for SZ in East Asian (EAS) population (P\u0026thinsp;\u0026lt;\u0026thinsp;5 \u0026times; 10⁻⁸) (\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e). Of the 21 reported lead SNPs, 19 variants were available in our dataset after quality control. We evaluated the association of these variants both in the overall SZ cohort and within each clinically defined subtype.\u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003eClustering analysis results\u003c/h2\u003e \u003cp\u003eIn the discovery stage, cluster analysis was conducted on 2,416 patients with SZ. Majority voting across 26 statistical indices suggested that either a two-cluster (supported by 7 indices) or three-cluster (supported by 7 indices) solution could be optimal. Both solutions demonstrated good stability (the \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\gamma\\:}_{c}\\)\u003c/span\u003e\u003c/span\u003e in two-cluster model were 0.95 and 0.97, and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\gamma\\:}_{c}\\)\u003c/span\u003e\u003c/span\u003e in three-cluster model were 0.79, 0.79, and 0.85). The two-cluster solution primarily separated patients by overall symptom severity (low vs. high), whereas the three-cluster solution revealed an additional subgroup characterized by predominant negative and disorganized symptoms. Results are presented in \u003cb\u003eSupplementary Fig.\u0026nbsp;1\u003c/b\u003e, with detailed comparisons provided in \u003cb\u003eSupplementary Results\u003c/b\u003e. Given that the three-cluster solution captured more clinically meaningful subgroups, offered greater interpretative value regarding symptom structure, and aligned with the recognized clinical subtypes of SZ(\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e), we selected this model for subsequent analyses.\u003c/p\u003e \u003cp\u003eThe symptom profiles of the three clusters are shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eA. Cluster-S (n\u0026thinsp;=\u0026thinsp;306, 12.7%) was characterized by the most severe symptoms across nearly all domains, except for negative symptoms. Cluster-N (n\u0026thinsp;=\u0026thinsp;711, 29.4%) was defined by predominant negative and disorganization symptoms. Cluster-L (n\u0026thinsp;=\u0026thinsp;1,399, 57.9%) exhibited generally mild symptom severity.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eClinical characteristics of the three subtypes are summarized in \u003cb\u003eSupplementary Table\u0026nbsp;5\u003c/b\u003e. Patients in Cluster-N had the highest mean age and the longest duration of illness compared to the other clusters. Both Cluster-S and Cluster-N were associated with poorer social functioning, as well as more severe manic, depressive, and anxiety symptoms, whereas patients in Cluster-L demonstrated comparatively preserved social functioning.\u003c/p\u003e \u003cp\u003eTo evaluate the reproducibility of the identified subtypes, we applied the same clustering procedure to the replication dataset (n\u0026thinsp;=\u0026thinsp;480). This analysis again yielded three stable subtypes, with \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\gamma\\:}_{c}\\)\u003c/span\u003e\u003c/span\u003eof 0.78, 0.79, and 0.76, indicating robust stability. The symptom structures and relative proportions of patients across the three subtypes were highly consistent with those observed in the discovery dataset (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eB). Spearman correlation analysis of WMW estimates further confirmed strong concordance for Cluster-L and Cluster-N (Spearman\u0026rsquo;s ρ\u0026thinsp;\u0026gt;\u0026thinsp;0.8), whereas concordance for Cluster-S was more moderate and did not reach the same high threshold (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eC).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003eDifferences of symptom networks among the identified subtypes\u003c/h2\u003e \u003cp\u003eThe network structures and centrality plots for the overall patient group and each subtype are shown in \u003cb\u003eSupplementary Fig.\u0026nbsp;2\u003c/b\u003e. Accuracy and stability analyses confirmed the robustness of all estimated networks (\u003cb\u003eSupplementary Figs.\u0026nbsp;3\u0026ndash;4\u003c/b\u003e). In the full patient sample, the nodes with the highest strength centrality were negative symptoms, positive symptoms, and mania. However, the most central nodes differed across subtypes. In Cluster-S, depressive symptoms, anxiety, and mania emerged as the most central nodes; in Cluster-N, positive symptoms, depressive symptoms, and anxiety were most central; and in Cluster-L, mania, negative symptoms, and positive symptoms were most central.\u003c/p\u003e \u003cp\u003eTo formally test whether the three subtypes exhibited distinct symptom structures, we conducted network comparison tests. Significant differences were observed in overall network structure, node centrality, and edge weights across subtypes. Detailed results are provided in the \u003cb\u003eSupplementary Results\u003c/b\u003e and \u003cb\u003eSupplementary Table\u0026nbsp;7\u003c/b\u003e.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003eGenome-wide association studies of all schizophrenia and its clinical subtypes\u003c/h2\u003e \u003cp\u003eWe conducted GWAS in the full SZ cohort and within each of the three subtypes, first in the discovery dataset and subsequently in the replication dataset. Principal component analysis confirmed that all participants were of East Asian ancestry (\u003cb\u003eSupplementary Fig.\u0026nbsp;5\u003c/b\u003e).\u003c/p\u003e \u003cp\u003eIn the discovery stage, 49 SNPs reached suggestive significance following LD-based clumping (20 in the full cohort, 6 in Cluster-S, 14 in Cluster-N, and 9 in Cluster-L). Of these, eight variants were successfully replicated (4 in the full cohort, 3 in Cluster-N, and 1 in Cluster-L).\u003c/p\u003e \u003cp\u003eMeta-analyses integrating results across both stages identified four independent loci that achieved genome-wide significance (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). The Manhattan and QQ plots for the meta-analyses are shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e. For Cluster-S, the genomic inflation factor (λGC\u0026thinsp;=\u0026thinsp;1.06) indicated only mild inflation, while LDSC analysis yielded an intercept of 1.05, suggesting that the inflation was attributable to polygenicity rather than confounding. Minimal inflation was observed in all other analyses (λGC\u0026thinsp;\u0026le;\u0026thinsp;1.02).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eGenome-Wide Association Results in Discovery, Replication, and Combined Samples.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"12\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\"\u0026times;\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\"\u0026times;\" class=\"colspec\" colname=\"c9\" colnum=\"9\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c10\" colnum=\"10\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c11\" colnum=\"11\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c12\" colnum=\"12\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eCluster\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eCHR\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eLead SNP\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eA1\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eNearest Gene\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c7\" namest=\"c6\"\u003e \u003cp\u003eDiscovery stage\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c9\" namest=\"c8\"\u003e \u003cp\u003eReplication stage\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"3\" nameend=\"c12\" namest=\"c10\"\u003e \u003cp\u003eMeta-analysis\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eOR\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eP\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c8\"\u003e \u003cp\u003eOR\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c9\"\u003e \u003cp\u003eP\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c10\"\u003e \u003cp\u003eOR\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c11\"\u003e \u003cp\u003eP\u003csub\u003emeta\u003c/sub\u003e\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c12\"\u003e \u003cp\u003eP\u003csub\u003eHET\u003c/sub\u003e\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAll\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e21\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ers2836330\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eG\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eAP001422.3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1.55\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026times;\" colname=\"c7\"\u003e \u003cp\u003e2.39 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;10\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e1.46\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026times;\" colname=\"c9\"\u003e \u003cp\u003e1.58 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;03\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e1.53\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e\u003cb\u003e1.64 \u0026times; 10\u003c/b\u003e\u003csup\u003e\u003cb\u003e\u0026minus;\u0026thinsp;12\u003c/b\u003e\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c12\"\u003e \u003cp\u003e0.640\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAll\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ers9876206\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eLRRC34\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1.43\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026times;\" colname=\"c7\"\u003e \u003cp\u003e1.31 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;07\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e1.35\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026times;\" colname=\"c9\"\u003e \u003cp\u003e1.25 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;02\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e1.40\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e\u003cb\u003e5.76 \u0026times;10\u003c/b\u003e\u003csup\u003e\u003cb\u003e\u0026minus;\u0026thinsp;09\u003c/b\u003e\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c12\"\u003e \u003cp\u003e0.677\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAll\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ers4240482\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eG\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eDENND1A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1.38\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026times;\" colname=\"c7\"\u003e \u003cp\u003e3.42 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;06\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e1.30\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026times;\" colname=\"c9\"\u003e \u003cp\u003e3.37 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;02\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e1.36\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e3.61 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;07\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c12\"\u003e \u003cp\u003e0.667\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAll\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e21\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ers220125\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eG\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eUMODL1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.75\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026times;\" colname=\"c7\"\u003e \u003cp\u003e9.87 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;06\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.77\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026times;\" colname=\"c9\"\u003e \u003cp\u003e2.14 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;02\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.75\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e6.37 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;07\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c12\"\u003e \u003cp\u003e0.863\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ers3767295\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eCNTN2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.60\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026times;\" colname=\"c7\"\u003e \u003cp\u003e7.77 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;09\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.68\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026times;\" colname=\"c9\"\u003e \u003cp\u003e1.73 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;02\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.62\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e\u003cb\u003e5.30 \u0026times; 10\u003c/b\u003e\u003csup\u003e\u003cb\u003e\u0026minus;\u0026thinsp;10\u003c/b\u003e\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c12\"\u003e \u003cp\u003e0.508\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e21\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ers970115\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eAP001422.3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1.53\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026times;\" colname=\"c7\"\u003e \u003cp\u003e1.43 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;06\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e1.70\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026times;\" colname=\"c9\"\u003e \u003cp\u003e8.21 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;04\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e1.57\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e\u003cb\u003e5.29 \u0026times; 10\u003c/b\u003e\u003csup\u003e\u003cb\u003e\u0026minus;\u0026thinsp;09\u003c/b\u003e\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c12\"\u003e \u003cp\u003e0.548\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ers56074668\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eRP11-73M7.6:COL16A1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.63\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026times;\" colname=\"c7\"\u003e \u003cp\u003e9.57 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;06\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.61\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026times;\" colname=\"c9\"\u003e \u003cp\u003e1.30 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;02\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.62\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e3.88 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;07\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c12\"\u003e \u003cp\u003e0.911\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eL\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e21\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ers2836330\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eG\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eAP001422.3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1.55\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026times;\" colname=\"c7\"\u003e \u003cp\u003e3.34 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;08\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e1.70\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026times;\" colname=\"c9\"\u003e \u003cp\u003e8.21 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;04\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e1.54\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e\u003cb\u003e8.13E-10\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c12\"\u003e \u003cp\u003e0.843\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"12\"\u003eVariant information and association statistics are shown for the most strongly associated SNP in each significant locus.\u003c/td\u003e\u003c/tr\u003e \u003ctr\u003e\u003ctd colspan=\"12\"\u003eAbbreviations: CHR: chromosome; SNP: single nucleotide polymorphism; A1: risk allele; OR: odds ratio; P: P values; P\u003csub\u003emeta\u003c/sub\u003e: P values for the meta-analysis of two-stage analysis; P\u003csub\u003eHET\u003c/sub\u003e: P values for Cochrane\u0026rsquo;s Q statistic\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eIn the meta-analysis of all patients, we identified two genome-wide significant loci (\u003cb\u003eTable\u0026nbsp;2\u003c/b\u003e). The first was located on chromosome 21 near the long non-coding RNA (lncRNA) gene \u003cem\u003eAP001422.3\u003c/em\u003e (lead SNP rs2836330, P\u0026thinsp;=\u0026thinsp;1.64 \u0026times; 10⁻\u0026sup1;\u0026sup2;, OR\u0026thinsp;=\u0026thinsp;1.53). The second was on chromosome 3 within the intronic region of \u003cem\u003eLRRC34\u003c/em\u003e (lead SNP rs9876206, P\u0026thinsp;=\u0026thinsp;5.76 \u0026times; 10⁻⁹, OR\u0026thinsp;=\u0026thinsp;1.40). Notably, the locus on chromosome 21 was also significant in Cluster-L (P\u0026thinsp;=\u0026thinsp;8.59 \u0026times; 10⁻\u0026sup1;\u0026sup1;, OR\u0026thinsp;=\u0026thinsp;1.54). Furthermore, a nearby variant (rs970115) within the same gene (\u003cem\u003eAP001422.3\u003c/em\u003e) achieved genome-wide significance in Cluster-N (P\u0026thinsp;=\u0026thinsp;5.29 \u0026times; 10⁻⁹, OR\u0026thinsp;=\u0026thinsp;1.57).\u003c/p\u003e \u003cp\u003eIn the subtype-specific meta-analyses, we identified one additional significant locus on chromosome 1, specific to Cluster-N. The lead SNP, rs3767295 (P\u0026thinsp;=\u0026thinsp;5.30 \u0026times; 10⁻\u0026sup1;⁰, OR\u0026thinsp;=\u0026thinsp;0.62), was in an intronic region of the \u003cem\u003eCNTN2\u003c/em\u003e gene. This SNP showed only nominal significance (P\u0026thinsp;\u0026lt;\u0026thinsp;0.05) in the other two subtypes and in the full patient analysis, with odds ratios closer to 1, indicating a markedly stronger effect in Cluster-N. Association results for all significant loci across subtypes are provided in \u003cb\u003eSupplementary Fig.\u0026nbsp;6\u003c/b\u003e, and the regional association plots are displayed in \u003cb\u003eSupplementary Figs.\u0026nbsp;7\u0026ndash;10\u003c/b\u003e.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003eReplication of previous EAS GWAS results\u003c/h2\u003e \u003cp\u003eWe further examined the associations of previously reported SZ risk variants identified in EAS population (\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e). Of the 19 variants available in our dataset, 11 showed nominally significant associations (P\u0026thinsp;\u0026lt;\u0026thinsp;0.05) in at least one analysis group. Importantly, the majority of these variants exhibited effect directions consistent with the original findings (\u003cb\u003eSupplementary Table\u0026nbsp;8\u003c/b\u003e). Within the combined patient group, 6 variants reached significance, of which only one was unique to this group, while the remaining 5 also showed associations in at least one subtype. Crucially, the subtype-based analyses revealed an additional 5 variants that were not significant in the overall patient group. These 5 variants were associated with only a single subtype, which 4 variants reached significance only in Cluster-S, and one only in Cluster-L.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003eGene-based analysis of all schizophrenia cases and its clinical subtypes\u003c/h2\u003e \u003cp\u003eGene-based analyses using MAGMA and H-MAGMA identified four significant genes in the overall patient group and twelve significant genes in Cluster-N after Bonferroni correction (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). In contrast, no genes reached statistical significance in Cluster-S or Cluster-L.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eIn the overall patient group, three protein-coding genes (\u003cem\u003eUMODL1\u003c/em\u003e, \u003cem\u003ePRDM15\u003c/em\u003e, and \u003cem\u003eLRRC34\u003c/em\u003e) and one non-coding RNA gene (\u003cem\u003eC21orf128\u003c/em\u003e) reached genome-wide significance. In Cluster-N, eight protein-coding genes, including \u003cem\u003eCNTN2\u003c/em\u003e, \u003cem\u003eDSTYK\u003c/em\u003e, \u003cem\u003eLRRN2\u003c/em\u003e, \u003cem\u003eNFASC\u003c/em\u003e, \u003cem\u003eRBBP5\u003c/em\u003e, \u003cem\u003eSLC2A13\u003c/em\u003e, \u003cem\u003eTMCC2\u003c/em\u003e, and \u003cem\u003eTMEM81\u003c/em\u003e, and four RNA genes, including \u003cem\u003eSNORD112\u003c/em\u003e, \u003cem\u003eTMCC2-AS1\u003c/em\u003e, \u003cem\u003eENSG00000228153\u003c/em\u003e, and \u003cem\u003eENSG00000240710\u003c/em\u003e were identified.\u003c/p\u003e \u003cp\u003eTo further investigate potential regulatory mechanisms underlying these associations, we performed eQTL mapping analyses (\u003cb\u003eSupplementary Table\u0026nbsp;9\u003c/b\u003e). In whole blood, rs2836330 and rs970115 were identified as eQTLs for \u003cem\u003eKCNJ15\u003c/em\u003e. In the dorsolateral prefrontal cortex (DLPFC), rs3767295 was associated with \u003cem\u003eCNTN2\u003c/em\u003e expression (P\u003csub\u003eFDR\u003c/sub\u003e = 2.42 \u0026times; 10⁻\u0026sup2;). Additionally, rs9876206 was associated with the expression of four genes in the DLPFC, including \u003cem\u003eMYNN\u003c/em\u003e (P\u003csub\u003eFDR\u003c/sub\u003e = 7.13 \u0026times; 10⁻\u0026sup3;), \u003cem\u003eLRRIQ4\u003c/em\u003e (P\u003csub\u003eFDR\u003c/sub\u003e = 3.76 \u0026times; 10⁻\u0026sup2;\u0026sup2;), \u003cem\u003eSAMD7\u003c/em\u003e (P\u003csub\u003eFDR\u003c/sub\u003e = 5.23 \u0026times; 10⁻\u0026sup3;), and \u003cem\u003eLRRC34\u003c/em\u003e (P\u003csub\u003eFDR\u003c/sub\u003e = 2.30 \u0026times; 10⁻\u0026sup1;\u0026sup2;).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003eGenome-wide PRS and pathway PRS analysis\u003c/h2\u003e \u003cp\u003eWe first calculated gPRS for eight psychiatric disorders across all subtypes and healthy controls. As shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eA, gPRS values for SCZ, BP were significantly elevated in all three subtypes relative to controls. Interestingly, the gPRS for MDD was significantly higher in Cluster-N (OR\u0026thinsp;=\u0026thinsp;1.20 [1.07\u0026ndash;1.35], P.adjust\u0026thinsp;=\u0026thinsp;0.02), but not in Cluster-S (OR\u0026thinsp;=\u0026thinsp;1.25 [1.07\u0026ndash;1.46], P.adjust\u0026thinsp;=\u0026thinsp;0.05) or Cluster-L (OR\u0026thinsp;=\u0026thinsp;1.15 [1.04\u0026ndash;1.28], P.adjust\u0026thinsp;=\u0026thinsp;0.06).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eWe next examined pPRS across four neurotransmitter pathways and ctPRS across six brain cell types, with results summarized in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eB and \u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eC. When comparing patients with controls, pPRS for glutamate were significantly elevated only in Cluster-L (OR\u0026thinsp;=\u0026thinsp;1.19 [1.07\u0026ndash;1.32], P.adjust\u0026thinsp;=\u0026thinsp;0.011) and Cluster-N (OR\u0026thinsp;=\u0026thinsp;1.27 [1.13\u0026ndash;1.43], P.adjust\u0026thinsp;=\u0026thinsp;0.001), but not in Cluster-S (OR\u0026thinsp;=\u0026thinsp;1.22 [1.04\u0026ndash;1.43], P.adjust\u0026thinsp;=\u0026thinsp;0.156), and pPRS for GABA were only elevated in Cluster-N (OR\u0026thinsp;=\u0026thinsp;1.19 [1.06\u0026ndash;1.34], P.adjust\u0026thinsp;=\u0026thinsp;0.032).\u003c/p\u003e \u003cp\u003eFor ctPRS, subtype-specific distinctions were also observed (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eC). Astrocyte-specific PRS (Ast-PRS) was significantly elevated only in Cluster-S (OR\u0026thinsp;=\u0026thinsp;1.29 [1.10\u0026ndash;1.52], P.adjust\u0026thinsp;=\u0026thinsp;0.035), whereas OPC-PRS was significantly elevated only in Cluster-N (OR\u0026thinsp;=\u0026thinsp;1.25 [1.11\u0026ndash;1.40], P.adjust\u0026thinsp;=\u0026thinsp;0.005). Oligodendrocyte PRS (Oli-PRS) was elevated in both Cluster-S (OR\u0026thinsp;=\u0026thinsp;1.31 [1.12\u0026ndash;1.53], P.adjust\u0026thinsp;=\u0026thinsp;0.012) and Cluster-N (OR\u0026thinsp;=\u0026thinsp;1.19 [1.06\u0026ndash;1.34], P.adjust\u0026thinsp;=\u0026thinsp;0.045). These findings suggest distinct, cell-type-specific genetic risk architectures across the identified subtypes.\u003c/p\u003e \u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eThis study identified three reproducible SZ subtypes based on PANSS symptom profiles. Subsequent genetic analyses, including subtype-specific GWAS, gene-based testing, and PRS analyses, revealed both unique risk loci and distinct genetic liability patterns across subtypes.\u003c/p\u003e \u003cp\u003eClinically, our analyses delineated three distinct subtypes: a general mild group (Cluster-L) with better overall functioning, a severe group (Cluster-S), and a group characterized by predominant negative and disorganized symptoms (Cluster-N). This three-cluster structure aligns closely with findings from previous studies that stratified SZ patients using PANSS scores, which reported comparable subtype in terms of both participant proportions and symptom profiles (\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e, \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e). These findings suggest that the three identified subtypes represent a robust and reproducible structure among SZ patients\u003c/p\u003e \u003cp\u003eSymptom network analysis further highlighted structural differences among subtypes. Notably, we observed a dissociation between symptom severity and network centrality. For example, although Cluster-N displayed the most severe negative symptoms, the negative symptom nodes showed relatively low centrality within the network. This suggests that, for these patients, negative symptoms may operate as an isolated and entrenched core feature of illness rather than being dynamically interconnected with other symptom domains. This interpretation is consistent with prior studies of similar patient subgroups(\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e) and may help explain the well-documented therapeutic resistance of negative symptoms in a subset of SZ patients(\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eOur analysis of the full SZ cohort identified two genome-wide significant loci: rs9876206 and rs2836330. The variant rs9876206 is located within \u003cem\u003eLRRC34\u003c/em\u003e, a gene encoding leucine-rich repeat containing, and play a role in DNA repair and telomere length regulation (\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e). However, the role of \u003cem\u003eLRRC34\u003c/em\u003e in SZ remains largely unexplored, underscoring the need for further mechanistic studies. The second significant variant, rs2836330, reached genome-wide significance in both the overall patient group and in Cluster-L. A secondary independent signal in the same region, rs970115, emerged as significant in Cluster-N. Both variants are eQTLs for \u003cem\u003eKCNJ15\u003c/em\u003e in whole blood. \u003cem\u003eKCNJ15\u003c/em\u003e encodes a potassium voltage-gated channel subunit and has previously been linked to Alzheimer\u0026rsquo;s disease(\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e), epilepsy(\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e), and Parkinson\u0026rsquo;s disease(\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e). Elevated expression of \u003cem\u003eKCNJ15\u003c/em\u003e in white blood cells has also been reported in patients with atypical depression and psychotic symptoms(\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e). However, its role in SZ has not been previously described, suggesting a novel avenue for investigation into its potential contribution to SZ pathophysiology.\u003c/p\u003e \u003cp\u003eWe also identified a genome-wide significant locus, rs3767295, located within \u003cem\u003eCNTN2\u003c/em\u003e, which was specific to Cluster-N. Notably, although previous large-scale GWAS have reported a suggestive association between SZ and another \u003cem\u003eCNTN2\u003c/em\u003e variant, rs11240341, which is in LD with rs3767295 in EAS population (R2\u0026thinsp;=\u0026thinsp;0.37) (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e, \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e, \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e), no study to date has identified a \u003cem\u003eCNTN2\u003c/em\u003e variant that reaches genome-wide significance for SZ. This underscores the utility of patient stratification for enhancing statistical power and uncovering subtype-specific risk loci. \u003cem\u003eCNTN2\u003c/em\u003e encodes contactin-2, a neural cell adhesion molecule involved in axon guidance, myelination, and neural development(\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e). Prior studies have reported reduced \u003cem\u003eCNTN2\u003c/em\u003e expression in the superior temporal gyrus of SZ patients(\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e). Conversely, a proteome-wide Mendelian randomization study found that higher genetically predicted CNTN2 protein levels in cerebrospinal fluid were associated with reduced SZ risk(\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e). Postmortem analyses further suggest that \u003cem\u003eCNTN2\u003c/em\u003e expression may vary by SZ subtype, showing increased expression in the amygdala of patients with disorganized SZ but decreased expression in those with paranoid SZ(\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e). Our findings are consistent with this complexity. In Cluster-N, characterized by prominent negative and disorganized symptoms, the rs3767295-C allele was associated with both lower SZ risk and reduced \u003cem\u003eCNTN2\u003c/em\u003e expression in the dorsolateral prefrontal cortex. These results suggest that lower \u003cem\u003eCNTN2\u003c/em\u003e expression may be protective in certain subtypes, but could also represent compensatory mechanisms in others. Thus, the relationship between \u003cem\u003eCNTN2\u003c/em\u003e and SZ risk appears non-linear and context-dependent, varying across brain regions and clinical subtypes.\u003c/p\u003e \u003cp\u003eWhen comparing gPRS across three severe psychiatric disorders, we observed a largely consistent pattern across the SZ subtypes, indicating a similar overall burden of common genetic risk. This is consistent with prior work showing that polygenic scores for SZ and BP did not differ significantly among psychosis biotypes(\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e). Notably, however, a clear exception emerged for MDD-gPRS: only Cluster-N demonstrated a significantly elevated MDD-gPRS relative to controls. This finding is particularly compelling given the marked clinical overlap between negative symptoms, such as anhedonia and avolition, and the core features of depression(\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e, \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e). Although a positive genetic correlation between SZ and MDD is well established(\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e), previous studies have not reported an association between MDD polygenic liability and specific clinical subgroups within psychosis(\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e, \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e). Our results suggest that Cluster-N may represent a subgroup of SZ patients with a distinct depression-related etiological pathway. This novel subtype-specific association highlights the potential of stratified approaches to reveal hidden genetic heterogeneity and warrants further investigation to clarify its biological mechanisms.\u003c/p\u003e \u003cp\u003eIn addition, we identified subtype-specific patterns in both pPRS and ctPRS. While certain pathways or cell types showed significantly elevated risk scores when all SZ patients were analyzed together, stratification into subtypes revealed distinct and more nuanced patterns. These differences were evident not only in comparisons between subtypes and healthy controls but also when contrasting individual subtypes against the remaining SZ patients. Such findings indicate that clinically defined subtypes may carry unique genetic liabilities tied to particular biological pathways and cellular processes. Unlike traditional gPRS, which aggregates genome-wide risk without functional context, pPRS provides a biologically informed framework by capturing genetic risk within specific gene sets or functional pathways. This improves interpretability and enhances its potential utility for disease stratification(\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e). Indeed, prior work has demonstrated that pPRS outperforms gPRS in predicting endophenotypes and is associated with psychosis-related biotypes, underscoring its value for uncovering hidden biological heterogeneity in complex psychiatric disorders(\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe present study has several notable strengths. We successfully identified and independently replicated clinically meaningful subtypes of SZ, providing strong evidence for their robustness. Moreover, by integrating subtype stratification with genetic analyses, we revealed distinct risk loci and biological patterns that were not detectable in traditional case\u0026ndash;control analyses.\u003c/p\u003e \u003cp\u003eSeveral limitations of this study should be acknowledged. The sample size, although substantial for a clinically detailed cohort, remains limited compared with recent large-scale case\u0026ndash;control GWAS of SZ. This limitation reflects the inherent difficulty of assembling cohorts that combine genomic data with comprehensive symptom assessments. In addition, the cross-sectional design provides only a static snapshot of symptomatology and cannot capture longitudinal trajectories or changes in subtype membership over time. Finally, our analyses were restricted to clinical symptom dimensions, without incorporating other important domains such as cognitive function, neuroimaging, or biomarkers, which could provide further refinement of subtype definitions. Future work leveraging larger, multi-ancestry, deeply phenotyped datasets, ideally with longitudinal follow-up, will be critical to developing more comprehensive and stable subtypes of SZ.\u003c/p\u003e"},{"header":"Conclusions","content":"\u003cp\u003eTaken together, our findings demonstrate that stratifying schizophrenia patients into clinically homogeneous subtypes provides a powerful framework for uncovering subtype-specific genetic architectures that would otherwise remain obscured in conventional case\u0026ndash;control analyses. The identification of novel loci, such as \u003cem\u003eCNTN2\u003c/em\u003e in the negative/disorganized subtype, alongside subtype-specific pathway and cell-type polygenic risks, underscores the biological validity of these subgroups. Importantly, this approach bridges clinical symptomatology with genetic mechanisms, offering a more nuanced understanding of the disorder\u0026rsquo;s heterogeneity. From a translational perspective, these insights raise the possibility that distinct subtypes may not only have different genetic underpinnings but could also differ in treatment response, disease trajectory, and prognosis. Future studies incorporating larger, multi-ancestry, and longitudinally phenotyped cohorts, as well as integrating multi-omics and neuroimaging data, will be crucial to refine these subtypes and link them to actionable biomarkers. Such efforts may pave the way toward precision psychiatry, where therapeutic strategies are tailored to biologically informed patient subgroups.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cdiv class=\"DefinitionList\"\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eAst\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eAstrocytes\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eBD\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003ebipolar disorder\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003ectPRS\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003ecell-type-specific polygenic risk scores\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eeQTL\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eexpression quantitative trait loci\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eEx\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eexcitatory neurons\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003egPRS\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003egenome-wide polygenic risk scores\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eGWAS\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003egenome-wide association study\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eHAMA\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eHamilton Rating Scale for anxiety\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eHAMD\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eHamilton Rating Scale for Depression\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eH-MAGMA\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eHi-C coupled MAGMA\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eIn\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003einhibitory neurons\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eLD\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003elinkage disequilibrium\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eMAGMA\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eMulti-marker Analysis of GenoMic Annotation\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eMDD\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003emajor depressive disorder\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eMic\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eMicroglia\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eOli\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eoligodendrocytes\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eOPC\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eoligodendrocyte progenitor cells\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003ePANSS\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003ePositive and Negative Syndrome Scale\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003ePGC\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003epsychiatric genomics consortium\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003epPRS\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003epathway-specific polygenic risk scores\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003ePRS\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003epolygenic risk scores\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eQC\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003equality control\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eSDSS\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eSocial Disability Screening Schedule\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eSZ\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eSchizophrenia\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eWMW\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eWilcoxon-Mann-Whitney\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eYMRS\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eYoung Mania Rating Scale\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003c/div\u003e"},{"header":"Declarations","content":" \u003cp\u003e \u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e \u003cp\u003e This study was approved by the Ethic Committee of West China Hospital, Sichuan University (2018\u0026thinsp;\u0026minus;\u0026thinsp;185) and the Ethic Committee of the Affiliate Mental Health Center, Zhejiang University School of Medicine (2025-019).\u003c/p\u003e \u003c/p\u003e \u003cp\u003e \u003cstrong\u003eConsent for publication\u003c/strong\u003e \u003cp\u003eNot applicable\u003c/p\u003e \u003c/p\u003e\u003cp\u003e \u003ch2\u003eCompeting interests\u003c/h2\u003e \u003cp\u003eThe authors declare no conflict of interest\u003c/p\u003e \u003c/p\u003e\u003ch2\u003eFunding:\u003c/h2\u003e \u003cp\u003eThis work was supported by the China Brain Project (STI2030-2021ZD0200404 and STI2030-2021ZD0200800 to TL); the Key\u0026middot;R\u0026amp;D\u0026middot;by Hangzhou\u0026middot;Science and\u0026middot;Technology\u0026middot;Bureau (20241203A14 to T.L.); the Zhejiang Clinovation Pride (CXTD202501053 to T.L.).\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eMZ, YZ, and TL conceived and designed the study; MZ, YZ, ZF, XW, HW, WW, HR, ML, QW, WD, and WG collected the data; MZ, YZ, and XD analyzed and interpretated the data; MZ made the figures, tables, and wrote the first draft; YZ, YZ, XL, and TL reviewed and revised. All authors read and approved the final paper.\u003c/p\u003e\u003ch2\u003eAcknowledgements\u003c/h2\u003e \u003cp\u003eWe thank all participating patients and their families for taking part in the study.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eThe data underlying this article will be shared on reasonable request to the corresponding author.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eOwen MJ, Legge SE. The nature of schizophrenia: As broad as it is long. Schizophr Res. 2022;242:109\u0026ndash;12.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTrubetskoy V, Pardi\u0026ntilde;as AF, Qi T, Panagiotaropoulou G, Awasthi S, Bigdeli TB, et al. Mapping genomic loci implicates genes and synaptic biology in schizophrenia. Nature. 2022;604(7906):502\u0026ndash;8.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOwen MJ, Legge SE, Rees E, Walters JTR, O'Donovan MC. Genomic findings in schizophrenia and their implications. Mol Psychiatry. 2023;28(9):3638\u0026ndash;47.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTaylor J, de Vries YA, van Loo HM, Kendler KS. Clinical characteristics indexing genetic differences in schizophrenia: a systematic review. Mol Psychiatry. 2023;28(2):883\u0026ndash;90.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDworkin RH, Lenzenweger MF. Symptoms and the genetics of schizophrenia: implications for diagnosis. Am J Psychiatry. 1984;141(12):1541\u0026ndash;6.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFanous AH, Zhou B, Aggen SH, Bergen SE, Amdur RL, Duan J, et al. Genome-wide association study of clinical dimensions of schizophrenia: polygenic effect on disorganized symptoms. Am J Psychiatry. 2012;169(12):1309\u0026ndash;17.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePardi\u0026ntilde;as AF, Smart SE, Willcocks IR, Holmans PA, Dennison CA, Lynham AJ, et al. Interaction Testing and Polygenic Risk Scoring to Estimate the Association of Common Genetic Variants With Treatment Resistance in Schizophrenia. JAMA Psychiatry. 2022;79(3):260\u0026ndash;9.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSada-Fuente E, Aranda S, Papiol S, Heilbronner U, Molt\u0026oacute; MD, Aguilar EJ, et al. Common genetic variants contribute to heritability of age at onset of schizophrenia. Transl Psychiatry. 2023;13(1):201.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJablensky A. Subtyping schizophrenia: implications for genetic research. Mol Psychiatry. 2006;11(9):815\u0026ndash;36.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWarren TL, Tubbs JD, Lesh TA, Corona MB, Pakzad SS, Albuquerque MD, et al. Association of neurotransmitter pathway polygenic risk with specific symptom profiles in psychosis. Mol Psychiatry. 2024;29(8):2389\u0026ndash;98.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHolliday EG, McLean DE, Nyholt DR, Mowry BJ. Susceptibility locus on chromosome 1q23-25 for a schizophrenia subtype resembling deficit schizophrenia identified by latent class analysis. Arch Gen Psychiatry. 2009;66(10):1058\u0026ndash;67.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDickinson D, Pratt DN, Giangrande EJ, Grunnagle M, Orel J, Weinberger DR, et al. Attacking Heterogeneity in Schizophrenia by Deriving Clinical Subgroups From Widely Available Symptom Data. Schizophr Bull. 2018;44(1):101\u0026ndash;13.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDwyer DB, Kalman JL, Budde M, Kambeitz J, Ruef A, Antonucci LA, et al. An Investigation of Psychosis Subgroups With Prognostic Validation and Exploration of Genetic Underpinnings: The PsyCourse Study. JAMA Psychiatry. 2020;77(5):523\u0026ndash;33.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMullins N, Forstner AJ, O'Connell KS, Coombes B, Coleman JRI, Qiao Z, et al. Genome-wide association study of more than 40,000 bipolar disorder cases provides new insights into the underlying biology. Nat Genet. 2021;53(6):817\u0026ndash;29.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNguyen TD, Harder A, Xiong Y, Kowalec K, H\u0026auml;gg S, Cai N, et al. Genetic heterogeneity and subtypes of major depression. Mol Psychiatry. 2022;27(3):1667\u0026ndash;75.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNarita A, Nagai M, Mizuno S, Ogishima S, Tamiya G, Ueki M, et al. Clustering by phenotype and genome-wide association study in autism. Transl Psychiatry. 2020;10(1):290.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLim K, Peh OH, Yang Z, Rekhi G, Rapisarda A, See YM, et al. Large-scale evaluation of the Positive and Negative Syndrome Scale (PANSS) symptom architecture in schizophrenia. Asian J Psychiatr. 2021;62:102732.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCharrad M, Ghazzali N, Boiteau V, Niknafs A, NbClust. An R Package for Determining the Relevant Number of Clusters in a Data Set. J Stat Softw. 2014;61(6):1\u0026ndash;36.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHennig C. Cluster-wise assessment of cluster stability. Comput Stat Data Anal. 2007;52(1):258\u0026ndash;71.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi L, Huang P, Sun X, Wang S, Xu M, Liu S, et al. The ChinaMAP reference panel for the accurate genotype imputation in Chinese populations. Cell Res. 2021;31(12):1308\u0026ndash;10.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWatanabe K, Taskesen E, van Bochoven A, Posthuma D. Functional mapping and annotation of genetic associations with FUMA. Nat Commun. 2017;8(1):1826.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang D, Liu S, Warrell J, Won H, Shi X, Navarro FCP et al. Comprehensive functional genomic resource and integrative model for the human brain. Science. 2018;362(6420).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ede Leeuw CA, Mooij JM, Heskes T, Posthuma D. MAGMA: generalized gene-set analysis of GWAS data. PLoS Comput Biol. 2015;11(4):e1004219.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSey NYA, Hu B, Mah W, Fauni H, McAfee JC, Rajarajan P, et al. A computational tool (H-MAGMA) for improved prediction of brain-disorder risk genes by incorporating brain chromatin interaction profiles. Nat Neurosci. 2020;23(4):583\u0026ndash;93.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRuan Y, Lin YF, Feng YA, Chen CY, Lam M, Guo Z, et al. Improving polygenic prediction in ancestrally diverse populations. Nat Genet. 2022;54(5):573\u0026ndash;80.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRuzicka WB, Mohammadi S, Fullard JF, Davila-Velderrain J, Subburaju S, Tso DR, et al. Single-cell multi-cohort dissection of the schizophrenia transcriptome. Science. 2024;384(6698):eadg5136.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLam M, Chen CY, Li Z, Martin AR, Bryois J, Ma X, et al. Comparative genetic architectures of schizophrenia in East Asian and European populations. Nat Genet. 2019;51(12):1670\u0026ndash;8.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLefort-Besnard J, Varoquaux G, Derntl B, Gruber O, Aleman A, Jardri R, et al. Patterns of schizophrenia symptoms: hidden structure in the PANSS questionnaire. Transl Psychiatry. 2018;8(1):237.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCowman M, Hodgekins J, Griffiths SL, Frawley E, O'Connor K, Fowler D et al. Cognitive and clinical profiles in first-episode psychosis and their relationship with functional outcomes. Br J Psychiatry. 2025:1\u0026ndash;8.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDemyttenaere K, Leenaerts N, Acsai K, Sebe B, Laszlovszky I, Barab\u0026aacute;ssy \u0026Aacute;, et al. Disentangling the symptoms of schizophrenia: Network analysis in acute phase patients and in patients with predominant negative symptoms. Eur Psychiatry. 2021;65(1):e18.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLeucht S, Chaimani A, Krause M, Schneider-Thoma J, Wang D, Dong S, et al. The response of subgroups of patients with schizophrenia to different antipsychotic drugs: a systematic review and meta-analysis. Lancet Psychiatry. 2022;9(11):884\u0026ndash;93.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePathare ADS, Džigurski J, Pujol-Gualdo N, Rukins V, Peters M, M\u0026auml;gi R, et al. A large-scale genome-wide association study on female genital tract polyps highlights role of DNA repair, cell proliferation, and cell growth. Hum Reprod. 2025;40(4):750\u0026ndash;63.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhou X, Chen Y, Mok KY, Zhao Q, Chen K, Chen Y, et al. Identification of genetic risk factors in the Chinese population implicates a role of immune system in Alzheimer's disease pathogenesis. Proc Natl Acad Sci U S A. 2018;115(8):1697\u0026ndash;706.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang S, Li Z, Ding X, Zhao Z, Zhang M, Xu H, et al. Integrative Analyses Identify KCNJ15 as a Candidate Gene in Patients with Epilepsy. Neurol Ther. 2022;11(4):1767\u0026ndash;76.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen X, Finol-Urdaneta RK, Chen M, Sykes AM, Gao B, Iqbal J, et al. Parkinson's disease-linked Kir4.2 mutation R28C leads to loss of ion channel function. J Physiol. 2025;603(12):3499\u0026ndash;518.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAhn I, Chang S, Choi S-H, Han J, Kim Y. Exploration of Novel Biomarkers through a Precision Medicine Approach Using Multi-omics and Brain Organoids in Patients with Atypical Depression and Psychotic Symptoms. bioRxiv. 2025:2025.04.21.649748.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi ZA-O, Chen J, Yu H, He L, Xu Y, Zhang D, et al. Genome-wide association analysis identifies 30 new susceptibility loci for schizophrenia. Nat Genet. 2017;49(11):1576\u0026ndash;83.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIkeda M, Takahashi A, Kamatani Y, Momozawa Y, Saito T, Kondo K, et al. Genome-Wide Association Study Detected Novel Susceptibility Genes for Schizophrenia and Shared Trans-Populations/Diseases Genetic Effect. Schizophr Bull. 2019;45(4):824\u0026ndash;34.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFan S, Liu J, Chofflet N, Bailey AO, Russell WK, Zhang Z, et al. Molecular mechanism of contactin 2 homophilic interaction. Structure. 2024;32(10):1652\u0026ndash;66.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRoussos P, Katsel P, Fau - Davis KL, Davis Kl Fau -, Bitsios P, Bitsios P, Fau - Giakoumaki SG, Giakoumaki Sg Fau - Jogia J, Jogia J, Fau - Rozsnyai K et al. Molecular and genetic evidence for abnormalities in the nodes of Ranvier in schizophrenia. Arch Gen Psychiatry. 2012;69(1):7\u0026ndash;15.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGu X, Dou M, Su W, Jiang Z, Duan Q, Cao B, et al. Identifying novel proteins underlying schizophrenia via integrating pQTLs of the plasma, CSF, and brain with GWAS summary data. BMC Med. 2022;20(1):474.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChang X, Liu Y, Hahn CG, Gur RE, Sleiman PMA, Hakonarson H. RNA-seq analysis of amygdala tissue reveals characteristic expression profiles in schizophrenia. Transl Psychiatry. 2017;7(8):e1203.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXia CA-O, Alliey-Rodriguez NA-O, Tamminga CA-O, Keshavan MA-OX, Pearlson GD, Keedy SK, et al. Genetic analysis of psychosis Biotypes: shared Ancestry-adjusted polygenic risk and unique genomic associations. Mol Psychiatry. 2025;30(6):2673\u0026ndash;85.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGuessoum SB, Le Strat Y, Dubertret C, Mallet J. A transnosographic approach of negative symptoms pathophysiology in schizophrenia and depressive disorders. Prog Neuropsychopharmacol Biol Psychiatry. 2020;99(Electronic):1878\u0026ndash;4216.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKrynicki CA-O, Upthegrove RA-O, Deakin JFW, Barnes TA-OX. The relationship between negative symptoms and depression in schizophrenia: a systematic review. Acta Psychiatr Scand. 2018;137(5):380\u0026ndash;90.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRomero CA-O, Werme JA-O, Jansen PA-O, Gelernter JA-O, Stein MA-O, Levey DA-O, et al. Exploring the genetic overlap between twelve psychiatric disorders. Nat Genet. 2022;54(12):1795\u0026ndash;802.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLegge SE, Cardno AG, Allardyce J, Dennison C, Hubbard L, Pardi\u0026ntilde;as AF, et al. Associations Between Schizophrenia Polygenic Liability, Symptom Dimensions, and Cognitive Ability in Schizophrenia. JAMA Psychiatry. 2021;78(10):1143\u0026ndash;51.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAhangari MA-O, Bustamante D, Kirkpatrick R, Nguyen TH, Verrelli BC, Fanous A et al. Relationship between polygenic risk scores and symptom dimensions of schizophrenia and schizotypy in multiplex families with schizophrenia. Br J Psychiatry.223(1):301\u0026ndash;8.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChoi SW, Garc\u0026iacute;a-Gonz\u0026aacute;lez JA-OX, Ruan YA-O, Wu HM, Porras CA-O, Johnson JA-O, et al. PRSet: Pathway-based polygenic risk score analyses and software. PLoS Genet. 2023;19(2):e1010624.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"bmc-medicine","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"bmed","sideBox":"Learn more about [BMC Medicine](http://bmcmedicine.biomedcentral.com/)","snPcode":"12916","submissionUrl":"https://submission.nature.com/new-submission/12916/3","title":"BMC Medicine","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Schizophrenia, Clinical subtypes, Genetics, Heterogeneity","lastPublishedDoi":"10.21203/rs.3.rs-8620966/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8620966/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003eClinical heterogeneity in schizophrenia (SZ) presents a significant challenge to genetic research, as diverse symptom profiles may stem from distinct genetic risk factors. Although previous studies have stratified patients into symptom-based subtypes, and preliminary evidence suggests the presence of distinct architectures, these findings remain limited. This study aimed to identify the clinically defined SZ subtypes and investigate the genetic architecture underlying different subtypes.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eIn a Chinese Han cohort of 2,410 SZ patients, we applied K-means cluster analysis to symptom profiles to identify clinical subtypes. The identified subtype structure was validated in an independent cohort of 480 patients. Subsequently, subtype-specific genome-wide association studies (GWAS) were conducted to identify genetic risk loci associated with individual subtypes.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eThree stable subtypes were identified: Cluster-L (low severity), Cluster-S (severe), and Cluster-N (predominant negative symptoms). Reproducibility of this classification was confirmed in the independent cohort. The three subtypes also exhibited significantly different symptom network structures. In GWAS analysis, A total of four genome-wide significant loci were detected, including a Cluster-N\u0026ndash;specific locus within the \u003cem\u003eCNTN2\u003c/em\u003e gene (lead SNP rs3767295, P\u0026thinsp;=\u0026thinsp;5.30E-10, OR\u0026thinsp;=\u0026thinsp;0.62). Gene-based analyses revealed additional risk genes unique to particular subtypes. Moreover, subtype-specific patterns emerged in both pathway-specific polygenic risk scores (pPRS) and cell type\u0026ndash;specific PRS (ctPRS).\u003c/p\u003e\u003ch2\u003eConclusions\u003c/h2\u003e \u003cp\u003eThese findings underscore the value of patient stratification in improving statistical power to detect subtype-specific risk loci. They further demonstrate that SZ patients with distinct symptom profiles harbor differential genetic liabilities involving diverse biological pathways and cell types.\u003c/p\u003e","manuscriptTitle":"Genetic Architecture of Schizophrenia Clinical Subtypes","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-02-05 09:13:16","doi":"10.21203/rs.3.rs-8620966/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2026-02-27T10:48:12+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-02-24T02:50:41+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-02-18T19:47:51+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"202548618194497478880776308245902001959","date":"2026-02-04T23:24:28+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"225623676621462361395853506791639276095","date":"2026-02-04T14:36:14+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-02-03T02:06:03+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2026-01-28T15:59:33+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-01-19T05:58:26+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-01-19T05:45:36+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Medicine","date":"2026-01-16T16:33:33+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bmc-medicine","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"bmed","sideBox":"Learn more about [BMC Medicine](http://bmcmedicine.biomedcentral.com/)","snPcode":"12916","submissionUrl":"https://submission.nature.com/new-submission/12916/3","title":"BMC Medicine","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"1d4d9539-dce7-4551-b93c-a07bd1a4a01d","owner":[],"postedDate":"February 5th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-05-09T03:25:08+00:00","versionOfRecord":[],"versionCreatedAt":"2026-02-05 09:13:16","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8620966","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8620966","identity":"rs-8620966","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}