A Long-read based Haplotype Panel Enhances Imputation and Discovery of Functional Small and Structural Variants

A Long-read based Haplotype Panel Enhances Imputation and Discovery of Functional Small and Structural Variants | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article A Long-read based Haplotype Panel Enhances Imputation and Discovery of Functional Small and Structural Variants Shaohua Fan, Tingting Gong, Yulu Zhou, Junfan Zhao, Yechao Huang, and 8 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6312842/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted You are reading this latest preprint version Abstract Haplotype reference panels are commonly used for genotype imputation in genome-wide association studies (GWAS). Although structural variations (SVs) are recognized as major contributors to human phenotypes, they are often excluded from GWAS analyses. Here, we integrate long-read-based and statistical methods to provide a comprehensive haplotype reference panel (Han-SV panel) incorporating 32,603,300 single nucleotide variants (SNPs), 3,180,227 small deletions and insertions and 172,569 SVs derived from 943 Han Chinese individuals. Our hybrid phasing approach had a 12.7-fold reduction in phasing error for small variants and 3.6-fold for SVs compared to conventional statistical phasing. This Han-SV panel enabled a more than two-fold in amount and four-fold in accuracy improvement of SV imputation compared to the expanded 1000 Genomes Project panel. Two GWASs using our panel-imputed variants identified 69 associated SVs and 101 previously unreported regions associated with skin-related and fingerprint phenotypes—substantially outperforming both short-read and SNP-array-based GWAS. This Han-SV panel offers a valuable resource for variant imputation and SV-included association studies to further uncover the novel phenotype associations and address critical gaps in missing heritability. An imputation server was provided for the use of the Han-SV panel ( https://www.biosino.org/svrp ). Biological sciences/Computational biology and bioinformatics Biological sciences/Genetics/Genomics Reference Panel Structural Variant Haplotype Phasing Long-Read Sequencing Genome-Wide Association Study Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Introduction Understanding the genetic architecture and uncovering genotype-phenotype associations is one fundamental goal of human genetics. With the significant advancement in cataloguing of genetic variants among different populations and the extensive phenotypic characterization of cohorts, numerous genetic variants have been identified to be associated with human phenotypes through genome-wide association studies (GWAS). However, a considerable fraction of the genetic variability underling complex phenotypes still remains unexplored, as current GWAS primarily focus on single nucleotide variants (SNVs) and small insertions and deletions (InDels), often overlooking large structural variations (SVs), leading to critical gaps in missing heritability 1 . Notably, SVs have been reported to be 50 times more likely to affect gene expression than SNVs and three times more likely to be associated with phenotypic traits in GWAS 2,3 . Therefore, the comprehensive discovery and functional interpretation of SVs are crucial for addressing missing heritability in complex traits 4,5 and identifying key disease risk factors 6,7 . Incorporating SVs into GWAS is both feasible and essential to fully capture the spectrum of genetic variation and its contributions to complex traits. In population-level GWAS studies, imputation on lower-cost microarray or low-coverage sequencing data using haplotype reference panels has now been the most practical and powerful strategy to identify associated variants with phenotypes. As such, large-scale efforts combining improved sequencing methodologies are now including a much larger and richer spectrum of SV haplotypes in haplotype reference panels. In 2015, the 1000 genomes project (1KGP) was the first to include SVs in its haplotype panel, utilizing low-coverage (~ 7.4X) whole-genome short-read sequencing (SRS) data of 2,504 individuals from 26 populations 1 . In 2022, 1KGP released an expanded reference panel, featuring high-coverage (30X) sequencing and including 602 trios 8 . In addition, great efforts have been made to develop population-specific reference panels using SRS data, such as Genome of the Netherlands (GoNL), Genome for Life Cohort (GCAT) and BioBank Japan (BBJ) developing SV-enriched haplotype reference panel for Dutch, Iberian and Japanese population, respectively 9–11 . However, these studies primarily rely on SRS technology, which can lead to an underrepresentation of SVs in reference panels. This is due to the short read length and the limitation of SRS to resolve highly repetitive and complex genomic regions 12,13 . Moreover, the large size of SVs—often exceeding 1 Mbp—contrasts with the relatively small haplotype blocks (typically 1–10 kb) generated through statistical phasing 14 , resulting in only a limited number of SV haplotypes being inferred from SRS data, often with suboptimal accuracy 15,16 . Although some haplotype panels, such as the expanded 1KGP and GoNL-SV panels, incorporate parental data to enhance phasing accuracy, SV haplotypes still rely heavily on statistical inference 8,9 . These limitations in SV detection and phasing ultimately reduce the imputation performance of haplotype panels, weakening the power of GWAS and fine-mapping causal variants. This highlights the need for haplotype panels that comprehensively capture the full spectrum of SV haplotypes in the human genome. Long-read sequencing technologies (LRS), capable of sequencing DNA fragments over 10 kb in length, far exceed the read lengths achievable with SRS, which are typically around 150 bp. LRS has been shown to significantly improve SV discovery 17–19 and provide access to genomic regions that are challenging to analyze with SRS data 12 . The comprehensive SV characterization based on LRS has offered new insights into SV functional interpretation, such as a rare DEL in PCSK9 found associated with cholesterol level 17 and causal SVs impacting GSDMD for bone density and WWP2 for height, weight, fat, craniofacial phenotypes, and innate immunity 19 . In addition, LRS has demonstrated its ability to significantly enhance haplotype phasing accuracy via direct read-based phasing without parents or large cohort data 20,21 . Building upon our previous work, which utilized whole-genome Oxford Nanopore long-read sequencing to generate a comprehensive SV set from 945 Han Chinese genomes 19 , this study provides a haplotype reference panel (Han-SV panel) to facilitate variant imputation for Chinese populations. Through hybrid phasing strategy combining long-read-based and statistical-based phasing methods, we achieve high phasing accuracy for both small variants and SVs in this Han-SV panel. Imputation performance evaluation of the Han-SV panel demonstrates its potential to significantly enhance SV imputation quality. We further conducted two SV-included GWASs, showing the value of our Han-SV panel to improve the power of GWAS and facilitate the identification of previously unreported variants associated with complex phenotypes and diseases. Results Haplotype panel construction and evaluation This study included 943 individuals with both whole genome SRS and LRS data from our previously published Han Chinese cohort 19 . From SRS data (~ 30X coverage), we identified 37,536,312 high-quality small variants, comprising 34,167,444 SNVs and 3,368,868 InDels, with a median of 3,987,826 SNVs and 931,557 InDels per sample (see Methods, Supplementary Fig. 1 ). Additionally, using Oxford Nanopore long-read sequencing data (100 individuals with ~ 30X and 843 with ~ 15X coverage), we identified 19,829 SVs per individual, with a median of 8,167 DELs, 10,923 INSs, 588 DUPs, and 155 INVs after quality control (see Methods , Supplementary Fig. 1 ). To generate a haplotype panel for the small variant and SV call sets, we applied a hybrid phasing workflow integrating long-read-based and statistical-based phasing methods (see Fig. 1 and Methods ). We first performed long-read-based co-phasing of SNVs, InDels and SVs using LongPhase 21 individually and merged the long-read phased haplotypes into one population-level haplotype set. Next, we applied SHAPEIT5 22 to statistically infer haplotypes for unphased variants, using the long-read phased variants as haplotype scaffold. We also included samples of HG002 from Genome in a Bottle (GIAB) Consortium 23,24 and LCL6 from the Chinese Quartet Genome Project 25 in our phasing workflow, with the purpose to access the phasing accuracy using their developed benchmark haplotype call sets. The GIAB developed HG002 benchmark haplotype-resolved set includes 2,869,654 bi-allelic heterozygous variants (2,360,987 SNVs, 492,454 InDels and 16,213 SVs) (see Supplementary Data 1 ). The haplotype-resolved variant set of LCL6 was generated from trio-based phasing with parents’ data (see Methods) , resulting in a total of 2,035,610 (2,029,241 SNVs and 6,369 SVs) bi-allelic heterozygous variants (see Supplementary Data 1 ). To access the variant phasing accuracy, published HG002 and LCL6 small variant sets were integrated into our SRS-called small variant set, which then was processed through the same phasing workflow. While raw ONT sequencing reads of HG002 and LCL6 were used for the same read alignment and SV calling, genotyping and phasing workflow as all samples in our Han Chinese cohort. The phasing accuracies were then assessed by comparing the phased set of HG002 and LCL6 to their benchmark haplotype sets as the reference truth set (see Methods ). Individual-level phasing and haplotype scaffold construction with long reads The generated long-read-based haplotype set included 36,265,725 small variants and 204,034 SVs in total, in which 21,515,772 (59.3%) small variants and 3,961 (1.9%) SVs were phased for all samples. When examining the phasing rates individually, we observed an average of 87% small variants (range 80% − 92%) and 84% SVs (range 64% − 87%) per-individual were phased using long-read-based phasing (see Supplementary Data 2 ). Among different types in SVs, DELs (92%, range 66–94%) and INSs (79%, range 62–83%) had higher average phasing rates compared to DUPs (60%, range 36–75%) and INV (67%, range 44–88%) (see Supplementary Data 2 ). We assessed the phasing accuracy of the small variant haplotype scaffold by calculating the switch error rate (SER) for HG002 and LCL6, using their benchmark haplotype sets as the reference truth set. In HG002, read-based co-phasing of SNVs, InDels and SVs using LongPhase achieved an average SER of 0.09% for small variants, while LCL6 showed a slightly higher average SER of 0.14% (see Fig. 2 A). We observed that multi-type variants co-phasing using LongPhase have lower SERs than read-based SNV-only phasing using WhatsHap (0.12% for HG002 and 0.21% for LCL6). Compared to previous SRS-based study of expanded-1KGP 8 (see Table 1 ), the SERs of small variants phasing in this study are close to the SERs of children (average 0.09%) from trio-based phasing, but notably lower than the reported SERs of unrelated individuals (average 0.79%) from statistical phasing. Table 1 Comparison of variant phasing accuracy with different methods Phasing method Small variant SER DEL flip rate INS flip rate Long-read-based multi-variant types co-phasing by LongPhase 0.09–0.14% 0.22–0.38% 0.34–0.69% Long-read-based phasing by WhatsHap 0.12–0.21% NA NA Hybrid phasing (statistical and long-read-based) 0.32–0.33% 0.36–0.52% 0.49–2.42% Statistical phasing on unrelated individuals 0.65–1.66% 1.80–4.75% 1.18–6.01% Trio-based phasing (Expanded-1KGP) 0.07–0.09% NA NA Hybrid phasing (statistical and trio-based, Expanded-1KGP) 0.79% 0.89% 0.24% Statistical phasing on unrelated individuals (BBJ-SV panel) NA 1.7% 3.5% The phasing accuracy of SVs in the haplotype scaffold was evaluated by calculating the flip error rate relative to the benchmark SV haplotype set of HG002 and LCL6. Comparing SVs in the haplotype scaffold and benchmark haplotype set, we defined two SVs as concordant if they are in the same type and breakpoints difference < 10 bp, accounting for breakpoint shift in SV reporting mainly due to micro-homology around SV breakpoint. We observed that 19% of the SVs (1,511 DELs and 1,577 INSs) in HG002 and 40% of the SVs (1,348 DELs and 1,135 INSs) in LCL6 in the benchmark sets as concordant in our haplotype scaffold (see Supplementary Fig. 2 ). Our analysis indicates a flip rate of 0.22% for DELs and 0.69% for INSs in HG002. For LCL6, we estimated flip rates of 0.38% and 0.34% for DELs and INSs, respectively (see Fig. 2 A). We found that long-read-based phasing of DELs had lower flip rates, while INSs exhibited greater flip rates than the rate reported by the expanded 1KGP (0.89% of DELs and 0.24% of INSs), where SV haplotypes were inferred with statistical phasing using a trio-based small variant haplotype scaffold (see Table 1 ). Additionally, our approach yielded lower flip rates for both DELs and INSs when compared to statistical phasing with unrelated samples (flip rate = 1.7% for DELs and 3.5% for INSs) as reported by BBJ-SV study 11 (see Table 1 ). Population-level haplotype phasing and panel development with hybrid approach With this hybrid phasing approach, the resulting integrated haplotype panel (Han-SV panel), consisting of 35,783,527 small variants and 172,569 SVs after quality control ( see Fig. 2 C and 2 D ). With the significant increase in the number of SVs per individual identified using LRS data in this study, we achieved nearly 5-fold more phased heterozygous SVs per individual in the Han-SV panel (16,998) compared to the expanded 1000 Genomes Project panel (3,462). The total number of SVs in our Han-SV panel is nearly two-fold more than that in the expanded-1KGP panel (97,862 across autosomes). The phasing accuracy of small variants resulting from the hybrid phasing approach was again evaluated based on the SERs of HG002 and LCL6. The SERs of small variants increased after statistical phasing (SER = 0.33% HG002 and SER = 0.32% for LCL6) compared to read-based phasing alone (see Fig. 2 A). Examining the SERs across different MAF bins (see Fig. 2 B), we found rare (MAF < 1%) small variants have lower phasing accuracy (1.5% for HG002 and 8.3% for LCL6), comparing to low-frequency (1% ≤ MAF < 5%) variants (0.78% for HG002 and 3.6% for LCL6) and common (MAF ≥ 5%) ones (0.23% and 0.26%). When compared against statistical phasing without the long-read-based scaffold, this hybrid phasing approach improved phasing accuracy by around 2-fold for LCL6 and 5-fold for HG002 on average and across different MAF bins (see Fig. 2 A and 2 B). The improvement in phasing accuracy using this hybrid phasing approach is also confirmed by comparing to the reported SERs by the expanded-1KGP (see Table 1 ), showing 2.4-fold decrease in SERs of small variants. We further evaluated the SV phasing accuracy by the hybrid phasing using flip error rate of HG002 and LCL6 (see Methods ) 8 . Overall, 19% of the SVs (1,384 DELs and 1,681 INSs) in HG002 and 25% of the SVs (571 DELs and 1,031 INSs) in LCL6 in the benchmark sets identified as concordant in our haplotype set and were compared to assess the flip error rate (see Supplementary Fig. 2 ). We observed that DEL phasing flip error rates increased by around 0.3% on average in HG002 and remained similar in LCL6 with additional statistical phasing, while INS flip error rates increased by 1.73% in HG002 and increased by 0.16% in LCL6, compared to long-read-based phasing alone (see Fig. 2 A). When compared against statistical phasing without the long-read-based scaffold, this hybrid phasing approach improved SV phasing accuracy around 5 to 9-fold for DELs and 2.4 to 5-fold for INSs (see Fig. 2 A and Table 1 ). Compared to published SV-included haplotype panels (see Table 1 ), we achieved lower flip rates of DEL compared to the expanded-1KGP study and for both DELs and INSs compared to BBJ-SV study 11 . Imputation performance evaluation of the haplotype panel To evaluate the imputation performance of our Han-SV panel, we utilized variants from 100 samples of this cohort on chromosome 20 with available WGS data as the imputation target. Variants from the remaining 843 samples were used to construct a test Han-SV panel using the hybrid phasing workflow (see Methods and Supplementary Fig. 3 ). Another two panels were employed for comparison: the expanded-1KGP panel and ChinaMAP. The expanded-1KGP is a widely used multi-ancestry reference panel, which includes high-quality SV haplotypes by high coverage SRS (30X) and trio data. While ChinaMAP 26 is one SNV imputation panel for the Chinese population, constructed with high coverage SRS data (41X). Overall, our panel imputed fewer total SNVs than expanded-1KGP and ChinaMAP (see Table 2 , Fig. 3 A and Supplementary Fig. 4 ). Assessing their imputation quality, defined by dosage r², revealed that our panel imputed a similar number of SNVs with dosage r² ≥ 0.1 to 0.9 (see Supplementary Fig. 4 ). Defining high-quality imputed variants as dosage r² ≥ 0.7, our panel imputed 163,565 SNVs, which is 12% more than the proportion of high-quality SNVs than the expanded-1KGP (159,611) and ChinaMAP (157,128). For InDels, although fewer high-quality InDels (15,637 vs. 45,799) were imputed compared to the expanded-1KGP panel, our panel achieved a comparable high-quality rate (24% vs 22%). In addition, our panel imputed 1.75-fold more total SVs and 4.3-fold more high-quality SVs, compared to the expanded-1KGP panel (see Table 2 , Fig. 3 B and Supplementary Fig. 4 ). As ChinaMAP is a SNV imputation panel, it has no InDels or SVs imputed (see Table 2 ). Table 2 Imputation quality with different reference haplotype panels Variant Type Imputation Count and Quality Expanded-1KGP ChinaMAP Test Han-SV panel SNV Total Count 1,365,653 1,353,675 690,259 High-quality (dosage r² ≥ 0.7) Count 159,611 157,128 163,565 F1 score 74% 72% 76% InDel Total Count 205,380 0 65,379 High-quality (dosage r² ≥ 0.7) Count 45,799 0 15,637 F1 score 40% NA 74% SV Total Count 2,203 0 3,874 High-quality (dosage r² ≥ 0.7) Count 292 0 1,265 F1 score 12% NA 53% To assess the accuracy of variant imputation, we compared the high-quality (dosage r² ≥ 0.7) imputed variants from each panel to their corresponding WGS-derived variants, and evaluated recall, precision, and F1 score across for each sample (see Methods , Fig. 3 C). Overall, our panel achieved a similar level of accuracy (72% − 76%), measured by F1 score, in SNV imputation, compared to the expanded-1KGP panel and ChinaMAP (see Table 2 ). As ChinaMAP is not able to impute InDels or SVs, we compared them to the expanded-1KGP panel (see Table 2 ) and observed a total of 34% increase of accuracy in InDel imputation and 41% increase in SV imputation by our panel. When examining different MAF bins, Han-SV and other two panels had higher imputation accuracy for common variants (MAF ≥ 5%), while lower for low-frequency (1% ≤ MAF < 5%) and rare (MAF < 1%) variants (see Fig. 3 C). Specifically, all three panels achieved nearly 80% imputation accuracy (F1 score) for common SNVs, while around 50% for low-frequency and only 10% for rare SNVs. For InDels, Han-SV panel had similar imputation accuracy as SNVs in different MAF bins, while the expanded-1KGP panel had lower F1 scores with 41% accuracy for common, 31% for low-frequency and 9.0% for rare InDel imputation. The accuracy of SV imputation is lower than other variant types in each MAF bins, with 54% for common, 25% for low-frequency and 12% for rare SVs by Han-SV panel. It should be noted our Han-SV panel imputed 4.3 times more high-quality SVs than the expanded 1KGP panel (see Fig. 3 B), therefore, leading to a significantly higher overall sensitivity (37.6% vs. 6.7%) (see Fig. 3 C). With the higher sensitivity in SV imputation, our panel is able to maintain the similar level of high precision for common SVs (> 85%) and low-frequency SVs (around 60%) as the expanded 1KGP panel. Moreover, a 13% higher precision (24% vs 11%) for rare SVs, compared to the expanded 1KGP panel can be achieved by our panel. Structural variant imputation and GWAS application on SNP array data To demonstrate the use and assess the value of Han-SV haplotype panel as an imputation resource for GWAS, we first performed variant imputation on SNP array data from a previously published GWAS study of fingerprint patterns (arch, loop, and whorl) of 10 fingers 27 based on 9,909 individuals from three Han Chinese cohorts (see Methods ). Overall, our panel imputed a comparable average number of SNVs (8,757,554 vs. 9,137,578) and fewer InDels (867,549 vs. 1,901,440) with dosage r² ≥ 0.7 in each cohort, while achieving approximately 2.3-fold more SVs (30,046 vs. 13,109) compared to the expanded-1000 Genomes Project (1KGP) panel (see Supplementary Data 3 ). We used the same significance threshold of GWAS (P-meta < 1.67 × 10⁻⁸) and compared our results with those from the original publication (see Supplementary Data 4, Methods ) 27 . A total of two SVs and 707 small variants (643 SNPs and 64 InDels) passed the significance threshold. Additionally, stepwise selection analysis using GCTA-COJO (v1.94.1) 28 identified 17 independent variants, which are all SNPs. Of the 18 previously reported fingerprint-associated loci 27 , we successfully replicated 15 of them. The remaining three loci were not replicated in our study due to low imputation qualities of the variants in these regions (dosage r² < 0.7) using Han-SV panel and were thus excluded from the GWAS analysis. Additionally, we identified two previously unreported association regions with lead SNPs achieving genome-wide significance (P-meta < 1.67 × 10⁻⁸) (see Supplementary Data 4 ). Although no SV reached genome-wide significant level (P-meta < 1.67 × 10⁻⁸) in the two previously unreported loci, we identified a 52 bp DEL on chromosome 18 (chr18: 55,586,143 − 55,586,194, GRCh38) within a previously identified locus that is marginally significantly associated with the fingerprint phenotype of the left little finger (D5L) (P-meta = 3.66 × 10⁻ 8 ) (see Fig. 4 A). The DEL is in linkage disequilibrium (r² = 0.68) with the lead variant, rs17089876 (P-meta = 1.53 × 10⁻ 9 ) in the original study 27 . The DEL is located in an enhancer region of TCF4 based on the annotation of the GeneHancer database, potentially disrupting its expression (see Fig. 4 B). Previous studies have shown that pathogenic mutations in TCF4 cause Pitt-Hopkins syndrome, which is characterized by persistent fetal digital pads 29–31 . Structural variant imputation and GWAS application on SRS data We also performed variant imputation on a SRS data called variant set, including 14,699,131 SNPs of an independent cohort with 1,016 samples of Han ancestry (see Methods ). A total of 70,949 SVs, 1,544,670 InDels and additional 7,536,153 SNPs were imputed with high-quality (dosage r² ≥ 0.7) (see Supplementary Data 5 ). We selected 145 continuous skin phenotypes and conducted a GWAS for each skin trait using both SRS-derived and imputed variants. Only high-quality imputed and SRS-called variants with MAF ≥ 1% (48,503 SVs and 8,158,252 SNPs/InDels) were included. As a result, three SVs and 669 small variants (617 SNPs and 52 InDels) reached genome-wide significance (P-value < 5 × 10⁻⁸) across 73 traits (see Methods ). We further performed stepwise selection analysis using GCTA-COJO (v1.94.1) 28 , identifying 219 independent variants (2 SVs and 217 SNPs/InDels) across 73 traits (see Supplementary Data 6 ). Defining association regions as ± 250kb around the independent variants (P-value < 5 × 10⁻⁸), we found that 104 regions (51.2%) in the present study overlap with the reported variants associated with skin phenotypes in previous GWAS 32 , while 99 regions (48.8%) were unreported. We found two association regions were identified by independent SVs using GCTA-COJO (v1.94.1) 28 —one previously reported and one unreported. Specifically, within a reported region, a 169 bp INS at chr8:143,072,208 (GRCh38) was found associated with the skin moisture phenotype (β = -3.381, P-value = 2.16 × 10⁻⁸) ( see Fig. 5 A and Supplementary Fig. 5A) . Functional annotation using AnnotSV 33 suggested that the 169 bp insertion causes a frameshift in two isoforms of LY6S (XM_054328425.1 and XM_054328424.1). The direct role of LY6S in skin moisture remains unclear. However, LY6 family proteins typically participate in cellular processes, including immune responses, which may indirectly affect skin health and moisture regulation 34,35 . In addition, we identified a 103 bp DEL (chr1:87,581,439 − 87,581,541, GRCh38), is significantly (β = -0.538, P-value = 3.15 × 10⁻ 10 ) associated with skin roughness in an unreported region ( see Fig. 5 B and Supplementary Fig. 5B) . The DEL is located in an intergenic region, and the nearest gene is LMO4 , as annotated by AnnotSV 33 . Previous studies have shown that LMO4 regulates keratinocyte proliferation and differentiation. Additionally, an overexpression of LMO4 has been observed in patients with psoriasis, a condition characterized by skin roughness 36–39 . Given our relatively small cohort (1,016 samples), we applied a more lenient significance threshold of 1 × 10⁻ 5 , uncovering an additional 64 SVs, 11,382 SNPs, and 1,769 InDels, resulting in a total of 3,951 independent variants (37 SVs and 3,914 SNPs/InDels) by GCTA-COJO (v1.94.1) 28 . Defining association regions as ± 250kb around independent variants, we observed 988 (55%) previously reported regions and 808 (45%) unreported regions compared to the reported skin related studies in NHGRI-EBI GWAS catalogue 28,32 (see Methods ). Overall, 36 association regions were identified through 37 independent SV signals, including 23 previously reported and 13 unreported regions (see Supplementary Data 7 ). Of the 35 independent signals identified with P-values ranging from 1 × 10⁻ 5 to 5 × 10⁻ 8 , functional annotation using AnnotSV revealed that nine SVs disrupt exons or enhancers of genes 33,40 (see Supplementary Data 7, Methods ). For example, we detected one 80 bp INS (chr15:101,185,413 − 101,185,413, GRCh38) associated with skin smoothness (β = -43.524, P-value = 6.85 × 10⁻ 6 ) which was located in the first exon of CHSY1 (XM_024449873.1), causing a frameshift mutation (see Fig. 5 C and Supplementary Fig. 5C ). Previous studies have shown that low expression of CHSY1 is associated with decreased skin thickness in aged skin 41 . Aged skin is characterized by the presence of mottled pigmentation, solar lentigines, rhytides (wrinkles), and xerosis (dryness), in contrast to the smoother, more even-toned appearance of younger skin. 41,42 . In addition, a 118 bp DEL (chr2:1,511,152-1,511,269, GRCh38) was positively associated with skin pigmentation based on Individual Typology Angle, ITA° (β = 1.451, P-value = 4.72 × 10⁻ 6 ) (see Fig. 5 D and Supplementary Fig. 5D ). Higher ITA° values correspond to lighter skin, while lower values correspond to darker skin 43 . The DEL is located in the intron 16 of TPO , overlapping an enhancer (GH02J001509) annotated by GeneHancer database that potentially regulates TPO and PXDN expression according to the Genehancer database 40 . Previous studies suggest that TPO is involved in thyroid hormone production, which influences hair follicle pigmentation 44,45 , and PXDN mutations can result in the white spot at the belly in mice 46 . Discussion In this study, we developed an SV-enriched haplotype reference panel for the Han Chinese population (Han-SV panel). Leveraging LRS and SRS data from 943 individuals, we combined long-read-based approaches with statistical phasing to accurately phase both small variants and SVs. This comprehensive haplotype set, alongside our integrated and highly accurate phasing method, represents a valuable resource for precise variant imputation and supports the inclusion of SVs in future GWAS. Based on Oxford Nanopore long-read sequencing data, we firstly demonstrated that direct long-read-based phasing of small variants has a similar high level of phasing accuracy (SERs = 0.09–0.14%) as SRS-read-based phasing (0.10%) 47 and trio-based phasing (0.07–0.09%) 8 . Without the extra cost on parents’ data collection and sequencing, read-based phasing is further more feasible than trio-based phasing to be applied in population-scale sequencing projects 48 . More importantly, long-read-based phasing facilitates the co-phasing of SVs with small variants, which is difficult with short-read-based phasing 14,21 . Using long-read-based co-phasing of SNVs, InDels and SVs in this study, we showed that the error rate of long-read-based SV phasing (0.22% − 0.69%) was lower than statistical-based phasing of unrelated individuals (1.7% − 3.5%) 11 and close to phasing with trio data (0.24% − 0.89%) 8 . The use of long-read sequencing data to improve SV phasing have also been shown by other studies, using long-read data only on patients 20 , integrating long reads with Hi-C data and/or single-strand genome sequencing data 3,13 . However, these methods have only been applied in individual-level or small cohorts with less than 50 individuals. In this study, we observed that approximately 90% of small variants and 80% of SVs can be phased based on direct long-read sequencing reads at the individual level. However, when variants from multiple individuals are combined into a single multi-sample variant set, only 60% of small variants and 1.9% of SVs are phased across all individuals. The lower population-level phasing rate of SVs may result from variations in long-read data alignment quality at the same SV site across different samples, as well as individual differences in breakpoints, both of which contribute to reduced phasing rates when merging SVs from multiple individuals. To facilitate population-level phasing and haplotype panel construction, we employed a hybrid phasing strategy that integrates both read-based and statistical approaches, by building a long-read-based scaffold for statistical phasing on unrelated individuals. Compared to the read-based method alone, this significantly improved the phasing rate at the population level—from 60–98.7% for small variants and from 1.9–84.6% for SVs—while only slightly increasing the switch error rate for small variants (by 2.7-fold, see Fig. 2 A) and the flip error rate for SVs (by 2.3-fold, see Fig. 2 A). Compared to statistical phasing without long-read-based scaffold, phasing error rates were reduced significantly for both small variants and SVs, with 12.7-fold reduction of SERs for small variants and 3.6-fold in flip rate for SVs (see Fig. 2 A). Compared to the published SV-included haplotype panels (see Table 1 ), both small variants and SVs were phased with error rates lower than statistical-phasing of unrelated individuals alone 8,11 . As such, this hybrid phasing strategy provides an efficient and feasible solution for future reference panel construction. As the commonly used resource in imputation and GWAS, the value of this newly constructed haplotype reference panel (Han-SV panel) in imputation was demonstrated with a similar number of high-quality imputed SNVs, compared to the commonly used multi-ancestry expanded-1KGP panel and Chinese population reference panel ChinaMAP. In contrast, Han-SV panel imputed 2.3 or 4.3-fold more SVs with high-quality, compared to the expanded-1KGP panel, on a generated test variant set and SNP array data from Li et al., 2022 27 . Evaluating the imputation accuracy of high-quality imputed variants, our Han-SV panel achieved similar or higher F1 scores for all variant types across different MAF bins, compared to ChinaMAP and the expanded-1KGP panel. In addition, imputation using this SV-enriched reference panel provides a more practical approach to include SVs in GWAS. In two GWASs of fingerprints and skin-related phenotypes implemented in this study, we not only identified SVs in previously known associated regions but also uncovered unreported associated regions when SVs were included in the GWAS analysis. For example, we identified a fingerprint trait associated DEL located in the prompter of TCF4 in a known associated region (see Fig. 4 A), while the lead SNP in this region is identified from SNP-only GWAS located in the intronic region (see Fig. 4 B). We also identified a previously unknown independent INS associated skin moisture phenotype, located in the exonic region of two isoforms of LY6S (see Fig. 5 A). Our findings include both independent associated SVs and SVs in linkage disequilibrium with previously reported associated variants, highlighting the value of SV-included GWAS to offer new insights into the genetic basis of complex traits in humans and the strong need of including SVs into future GWAS. In conclusion, we presented a Han Chinese haplotype reference panel, including both small variants and SVs, and its high imputation performance allows GWAS to be implemented without additional sequencing costs. This haplotype panel therefore provided a practical and cost-effective solution for the robust analysis of common SVs in the Chinese population, which will strongly advance our knowledge of the genetic underpinnings of disease and complex phenotypes. Materials and Methods LRS and SRS data generation This study included 943 Han Chinese individuals, recruited from Zhengzhou City in central China. DNA was extracted from whole blood samples, as described previously 19 . In brief, short-read WGS was performed using the BGI DNBSEQ-T7 platform and long-read WGS was performed with Oxford Nanopore Technologies on the PromethION platform. Short-read sequencing reads were aligned to the GRCh38 reference genome using BWA-MEM mode (v0.7.17) with default parameters 49 , resulting in > 30X depth of coverage per sample. Oxford Nanopore sequencing reads were aligned to GRCh38 reference gnome using NGMLR (v0.2.7) with default parameters 50 . The LRS mean depth of coverage of 100 samples was 28X (range: 22X − 34X) and 843 samples was 14X (range: 10X − 26X). Genome-wide variant calling and genotyping Small variants (including SNV and InDel) were identified based on SRS data using Genome Analysis Toolkit (v4.1.8.1) following the best practice for germline variant calling 51,52 . Structural variants were called individually based on ONT data using sniffles (v2.2) 53 , requiring output names of all supporting reads for each SV (--output-rnames). The tandem repeat annotations for the human reference genome (sniffles accompanied annotation file: human_GRCh38_no_alt_analysis_set.trf.bed) was provided as an additional input to improve SV calling in the repetitive regions in GRCh38. We only included autosomal SVs with PASS in the FILTER field in VCF in the further analyses. We further excluded TRAs, SVs with size less than 50 bp, SVs in telomere, centromere or gap regions or very large SVs with length greater than one third of the chromosome 12 . We implemented SV genotyping individually using LRcaller (v1.0) with default parameters 17 . To obtain higher accuracy in genotyping, we filled the sequences in GRCh38 affected by SV in the REF and ALT fields before LRcaller genotyping to increase genotype accuracy. Benchmarking samples To evaluate the accuracy of phasing methods, we included samples of HG002 from the Genome in a Bottle (GIAB) Consortium 23,24 , and LCL6 from the Chinese Quartet Genome Project 25 , both of which have established benchmark small variant and SV call sets. The benchmark small variant set (NISTv4.2.1) of HG002 was downloaded from GIAB ( https://ftp-trace.ncbi.nlm.nih.gov/ReferenceSamples/giab/release/AshkenazimTrio/HG002_NA24385_ son/NISTv4.2.1/GRCh38/HG002_GRCh38_1_22_v4.2.1_benchmark.vcf.gz ). The benchmark small variant and SV sets of LCL6 and her parents (LCL7 and LCL8) were downloaded from the Chinese Quartet Genome Project ( http://chinese-quartet.org ). The benchmark haplotype-resolved small variant and SV set (CMRG_v1.00) of HG002 based on GRCh38 were downloaded from GIAB ( https://ftp-trace.ncbi.nlm.nih.gov/ReferenceSamples/giab/release/AshkenazimTrio/HG002_NA24385_son/CMRG_v1.00/GRCh38/Additional file 2:Files/HG002v11-align2-GRCh38/HG002v11-align2-GRCh38.dip.vcf.gz), which was used as truth set for phasing accuracy evaluation. To generate a haplotype-resolved variant set of LCL6, we implemented WhatsHap (v1.4) phasing in pedigree mode 54 based on downloaded benchmark small variant and SV call sets of the trio, respectively. To include HG002 and LCL6 in our phasing pipeline, their ONT sequencing data were obtained as follows. The ONT sequencing data in FASTQ format of HG002 were downloaded from GIAB ( https://ftp-trace.ncbi.nlm.nih.gov/ReferenceSamples/giab/data/AshkenazimTrio/HG002_NA24385_ son/UCSC_Ultralong_OxfordNanopore_Promethion/GM24385_1.fastq.gz ), and mapped to GRCh38 using NGMLR (v0.2.7) with default parameters 50 . The average depth of coverage of HG002 was 40X. The ONT sequencing alignment data (aligned by NGMLR) of LCL6 and her parents (LCL7 and LCL8) were downloaded from the Chinese Quartet Genome Project ( ftp://download.big.ac.cn/gsa-human/HRA001859/HRR796651 ). Due to the coverage of these samples near 100X, we downsampled them to 27X, 38X, and 39X respectively. The SV calling and genotyping of HG002 and LCL6 were implemented the same as described above based on ONT data. Integrating long-read-based and statistical-based phasing for small variants and SVs The benchmark small variant datasets from HG002 and LCL6 were merged with the small variant set of our Han Chinese cohort using the bcftools merge module (v1.20) 55 to create a comprehensive multi-sample variant set. Before phasing, the following filters were applied: genotype missing rate 1 × 10⁻ 10 , and allele count > 0. The resulting multi-sample VCF was then split into individual VCFs to enable read-based phasing at the individual level. Individual-level long-read-based phasing Small variants and SVs for each sample in our Han Chinese cohort, as well as for HG002 and LCL6, were co-phased using LongPhase (v1.5.1) with the sequencing platform specified (--ont) and parameters provided for co-phasing InDels and SVs (--indels and --sv-file) 21 . To compare the performance of the read-based co-phasing method with SNV-only phasing, long-read-based phasing of small variants for HG002 and LCL6 was also performed using WhatsHap phase (v1.4) with default settings 54 . All phased small variant VCF per individual from longphase were merged as one multi-sample VCF file using bcftools merge (v1.20) 55 . We used Truvari collapse (v4.1.0) 56 , a tool designed for SV merging, to combine the phased SVs from the individual SV VCFs, requiring minimum percentage of allele sequence and size similarity as 0.5, and other comparison threshold arguments as default. Population-level statistical phasing Several quality control measures for SVs were applied before population-level statistical phasing. This includes removing variants lacking exact sequences in the REF or ALT fields of the VCF, such as INSs without resolved inserted sequences. To prevent a single variant from being classified as both an InDel and an SV in the same sample, InDels larger than 50 bp were excluded. The multi-sample SV set was re-genotyped using LRCaller (v1.0) with default parameters 17 , before statistical phasing to reduce the genotype missing rate caused by SV merging at individual level. Multi-sample set of phased small variants and SVs, as generated through long-read-based phasing (as described above), were then merged using bcftools concat (v1.20) 55 to construct a haplotype scaffold. Statistical phasing was performed using this scaffold with SHAPEIT5 phase_common (v5.1.1), employing the genetic map files for GRCh38 provided by SHAPEIT5 22 . The resulting haplotype set was subsequently filtered using the following criteria: genotype missing rate 1 × 10⁻ 10 , and allele count > 0. Haplotype phasing performance evaluation The phasing accuracy of the small variant and structural variant (SV) haplotype sets for HG002 and LCL6 was evaluated by comparing them to their respective benchmark haplotype sets. For small variants, phasing accuracy was quantified using the switch error rate (SER), calculated with the WhatsHap compare module (v1.4) 54 . For SVs, phasing accuracy was measured by the flip error rate, which represents the proportion of SVs with phase flips based on SV-flanking-SNV pairs, as previously described by Byrska-Bishop et al., 2022 and Kosugi et al., 2023 8,11 . Briefly, we calculated the switch error for each phased SV and its nearest small variants on both sides, comparing our phased set to the benchmark haplotype set. Concordant SVs between our phased and benchmark sets were defined as having matching SV types and breakpoints within 10 bp. Concordant small variants identification was computed using bcftools isec (v1.20) to identify small variants with exact same position in phased and benchmark sets 55 . A flip error was defined as the presence of two switch errors 8,11 . Imputation performance evaluation We generated a test dataset by randomly selecting 100,000 SNVs from chromosome 20 across 100 randomly selected samples from the Han Chinese cohort. Imputation was performed using Beagle (v5.4) with default parameters 57 , utilizing the haplotype set from the remaining 843 samples as the test Han-SV panel (see Supplementary Fig. 3 ). To compare the imputation performance to other imputation panels, the expanded-1KGP panel and ChinaMAP were also used as reference to impute the test dataset. The expanded-1KGP panel was obtained from http://ftp.1000genomes.ebi.ac.uk/vol1/ftp/data_collections/1000G_2504_high_coverage/working/20220422_3202_phased_SNV_indel_SV , and imputation was performed using Beagle (v5.4) with default parameters 57 . The ChinaMAP imputation was performed using its imputation server ( http://www.mbiobank.com/imputation/ ) with default settings 26 . To estimate imputation accuracy, the imputed variants on chromosome 20 of the 100 samples were compared to the variant sets derived from WGS data as truth sets (see Supplementary Fig. 3 ). For small variants, we compared the imputed genotypes of each sample with the ones derived from the SRS data of the same sample. The imputed true positive small variants were defined as having the same genomic position and reference and alternative alleles as that in SRS data. For SVs, we compared the imputed SVs of each sample with the ones derived from LRS data of the sample sample. The individual LRS-derived SV set of the 100 samples were merged using truvari collapse (v4.1.0) 56 , requiring minimum percentage of allele sequence and size similarity as 0.5 and other comparison threshold parameters as default. The 100-sample SV set was re-genotyped using LRCaller (v1.0) with default parameters 17 to reduce the genotype missing rate caused by SV merging at individual level. The imputed true positive SVs were defined as having the same SV type and breakpoints within 200 bp as that in the LRS-derived SV set. When assessing imputation accuracy for different MAF bins, both imputed variant and truth set were required in the same MAF bin. Recall was calculated as the fraction of true positives in the truth set and precision was calculated as the fraction of true positives in the imputed set. F1 score was calculated as the harmonic mean of recall and precision. Imputation and GWAS using SNP array data SNP array data and ten fingerprint phenotypes of 9,909 individuals, from three Han Chinese cohorts: the Taizhou Longitudinal Study (TZL, n = 2,961), the National Survey of Physical Traits (NSPT, n = 2,679) and the Jidong cohort Study (JD, n = 4,269), were obtained from Li et al., 2022 27 . The SNP set was filtered using the following criteria: genotype missing rate 1 × 10⁻ 10 and allele count > 0. Since the SNP array data was based on GRCh37, we converted the genomic coordinates to GRCh38 using LiftoverVcf from Picard tool (v2.27.5) ( http://broadinstitute.github.io/picard ). Imputation was performed using our Han-SV haplotype reference panel with Beagle (v5.4) with default parameters 57 . To compare the imputation performance of Han-SV panel and the expanded-1KGP panel on SNP array data, the expanded-1KGP panel was also used as reference to impute TZL SNP array data on chromosome 20 using Beagle (v5.4) with default parameters 57 . Quality control was applied before GWAS, including removing variants with low imputation quality (dosage r 2 < 0.7 from Beagle results) and rare variants with minor allele frequency 0.2. The genome-wide association analyses on ordinal phenotypes (coded as 0, 1, and 2 for arch, loop, and whorl, respectively) of ten fingers were performed using PLINK (v2.0) 58 , using multiple linear regression model of additive allelic effects with additional 4 genetic principal components (PCs), 5 genetic PCs, and 5 genetic PCs as covariates in TZL, NSPT, and JD cohorts, respectively as previously described in the original study 27 . We used METAL (version 2011-03-25) 59 to perform fingerprint meta-analyses across the three independent cohorts and selected genome-wide significant variants (P-meta < 1.67 × 10⁻ 8 ) after multiple-testing adjustment 27 for further analysis, and we further defined the marginally significance as 1.67 × 10⁻ 8 < P-meta < 3.33 × 10⁻ 6 . Imputation and GWAS using SRS-derived variants An independent cohort including 1,016 Han Chinese individuals were recruited from Shanghai city in China. As previously described 60 , short-read WGS was performed using the BGI-DNBSEQ-T7 platform. For each sample, we obtained at least 90 Gb data. Short-read sequencing reads were aligned to the GRCh38 reference genome using BWA-MEM mode (v0.7.17) with default parameters 49 . Small variants (SNV and InDel) were identified using Genome Analysis Toolkit (v4.1.7.0) following the best practice for germline variant calling 51,52 . Only biallelic SNVs with PASS in FILTER field of VCF were included for imputation. Imputation was performed using the Han-SV panel with Beagle (v5.4) with default parameters 57 . The imputed variants were firstly filtered based on imputation quality (dosage r 2 ≥ 0.7), provided by Beagle software 57 . The imputed and called variants were also filtered using PLINK (v1.90 beta 6) 58 specifying minimum MAF (–maf 0.01) and maximal missing rate (–geno 0.2). Genome-wide association analysis was performed for 145 skin-related traits ( see Supplementary Data 8 ) in 1,016 samples on the filtered SV and SNP/InDel markers using the Linear Mixed-Model Association program (EMMAX) 61 with sex, age, and first 6 PCs as covariates. We defined genome-wide significance as P-value < 5 × 10⁻ 8 and the suggestive significance as 5 × 10⁻ 8 < P-value < 1 × 10⁻ 5 . Variant annotation of significant loci in GWAS To identify independent variants, a stepwise selection was conducted using GCTA-COJO (v1.94.1) with default parameters 28 . We defined association regions as genomic regions within 250kb up- and down-stream of the independent variants, and merged the overlapped genomic regions to obtain association regions 27 . Variants within association regions and with LD r 2 ≥ 0.3 of the independent variants were defined as candidate variants 62,63 . All candidate SVs were annotated using AnnotSV (v3.4) with default settings 33 . All candidate SNPs/InDels were annotated by SnpEff (v5.2a) 64 , using the GRCh38.105 database as canonical transcripts (-canon). To annotate variants in non-coding regions of the genome, regulatory region annotations were obtained from the GeneHancer database 40 . SNPs and InDels were identified overlapping regulatory elements if presenting within a ± 200 bp flanking region, using bedtools intersect 65 . SVs were considered to overlap regulatory regions if either breakpoint was within 200 bp of the regulatory elements. A comparison to the association regions in the EBI GWAS category We downloaded the association file from the NHGRI-EBI GWAS catalogue (release version r2023-06-17) 32 and extracted results for skin redness-related phenotypes (hemoglobin measurement), skin disease-related phenotypes (psoriasis, melanoma, keratinocyte cancers), pigmentation-related phenotypes (hair, skin, and eye color), and radiation response (see Supplementary Data 9 ). Association regions of skin phenotypes identified in this study were defined as previously reported regions if they overlapped with the variants reported in the NHGRI-EBI GWAS catalogue 32 . Declarations Ethics approval and consent to participate The study complied with all relevant regulations for working with human subjects in China. The Ethics Committee of the School of Life Sciences, Fudan University, Shanghai, China approved the study. Participants were recruited to a project studying physical anthropology diversity in China funded by the Ministry of Science and Technology of the People’s Republic of China (2015FY111700). Informed consents were approved by all participants. Consent for publication Not applicable. Availability of data and materials The use of Han-SV haplotype reference panel is available through the developed imputation server (https://www.biosino.org/svrp), that can be used to impute both small and structural variants. Acknowledgements We thank Dr. Marta Byrska-Bishop for providing the detailed guide to calculate SV phasing flip error rate, and the help from Professor Leming Shi and Dr. Luyao Ren for accessing the Zhonghua Quartet Project data. We also like to acknowledge the support from the Human Phenome Data Center at Fudan University. Funding This work was supported by grants from the National Key R&D Program of China (2024YFE011610), Noncommunicable Chronic Diseases-National Science and Technology Major Project (2024ZD0533101, and 2024ZD0533100), National Natural Science Foundation of China (grant No. 32370686), and Shanghai Municipal Science and Technology (grant No. 2023SHZDZX02A13) to S.F. and the 111 Project (B13016) to S.F and L.J.. T.G was supported by the Office of China Postdoctoral Council International Postdoctoral Exchange Program Fellowship (YJ20210053). Author contributions: S.F. and L.J. conceived the study. T.G. performed reference panel construction and the evaluation. Y.Z. performed the GWAS applications. Y.H., J.Z. and J.G. provided additional supports in Oxford Nanopore sequencing data processing and J.L. provided supports in GWAS analysis. J.L., Q.P. and S.W. contributed to generating the SNP array and fingerprint phenotype data. T.G. and Y.Z. did the data analysis and interpretation. H.C., L.Y. and G.Z. developed the imputation server. T.G., Y.Z. and S.F., wrote the manuscript. S.F. and L.J. supervised the study. All authors reviewed the manuscript. Conflict of interest statement : None declared. References Sudmant, P. H. et al. An integrated map of structural variation in 2,504 human genomes. Nature 526 , 75–81 (2015). Chiang, C. et al. The impact of structural variation on human gene expression. Nat. Genet. 49 , 692–699 (2017). Chaisson, M. J. P. et al. Multi-platform discovery of haplotype-resolved structural variation in human genomes. Nat. Commun. 10 , 1784 (2019). Mukamel, R. E. et al. Protein-coding repeat polymorphisms strongly shape diverse human phenotypes. Science 373 , 1499–1505 (2021). Wheeler, M. M. et al. Whole genome sequencing identifies structural variants contributing to hematologic traits in the NHLBI TOPMed program. Nat Commun 13 , 7592 (2022). Billingsley, K. J. et al. Genome-Wide Analysis of Structural Variants in Parkinson Disease. Ann Neurol 93 , 1012–1022 (2023). Kaivola, K. et al. Genome-wide structural variant analysis identifies risk loci for non-Alzheimer’s dementias. Cell Genom 3 , 100316 (2023). Byrska-Bishop, M. et al. High-coverage whole-genome sequencing of the expanded 1000 Genomes Project cohort including 602 trios. Cell 185 , 3426–3440.e19 (2022). Hehir-Kwa, J. Y. et al. A high-quality human reference panel reveals the complexity and distribution of genomic structural variants. Nat. Commun. 7 , 12989 (2016). Valls-Margarit, J. et al. GCAT|Panel, a comprehensive structural variant haplotype map of the Iberian population from high-coverage whole-genome sequencing. Nucleic Acids Res. 50 , 2464–2479 (2022). Kosugi, S. et al. Detection of trait-associated structural variations using short-read sequencing. Cell Genom. 3 , 100328 (2023). Audano, P. A. et al. Characterizing the major structural variant alleles of the human genome. Cell 176 , 663–675.e19 (2019). Ebert, P. et al. Haplotype-resolved diverse human genomes and integrated analysis of structural variation. Science 372 , eabf7117 (2021). Glusman, G., Cox, H. C. & Roach, J. C. Whole-genome haplotyping approaches and genomic medicine. Genome Med. 6 , 73 (2014). Zhang, X., Wu, R., Wang, Y., Yu, J. & Tang, H. Unzipping haplotypes in diploid and polyploid genomes. Comput. Struct. Biotechnol. J. 18 , 66–72 (2020). Garg, S. Computational methods for chromosome-scale haplotype reconstruction. Genome Biol. 22 , 101 (2021). Beyter, D. et al. Long-read sequencing of 3,622 Icelanders provides insight into the role of structural variants in human diseases and other traits. Nat. Genet. 53 , 779–786 (2021). Wu, Z. et al. Structural variants in the Chinese population and their impact on phenotypes, diseases and population adaptation. Nat. Commun. 12 , 6501 (2021). Gong, J. et al. Long-read sequencing of 945 Han individuals identifies structural variants associated with phenotypic diversity and disease susceptibility. Nat Commun 16 , 1494 (2025). Cretu Stancu, M. et al. Mapping and phasing of structural variation in patient genomes using nanopore sequencing. Nat. Commun. 8 , 1326 (2017). Lin, J.-H., Chen, L.-C., Yu, S.-C. & Huang, Y.-T. LongPhase: an ultra-fast chromosome-scale phasing algorithm for small and large variants. Bioinformatics 38 , 1816–1822 (2022). Hofmeister, R. J., Ribeiro, D. M., Rubinacci, S. & Delaneau, O. Accurate rare variant phasing of whole-genome and whole-exome sequencing data in the UK Biobank. Nat. Genet. 55 , 1243–1249 (2023). Wagner, J. et al. Benchmarking challenging small variants with linked and long reads. Cell Genom. 2 , 100128 (2022). Wagner, J. et al. Curated variation benchmarks for challenging medically relevant autosomal genes. Nat. Biotechnol. 40 , 672–680 (2022). Ren, L. et al. Quartet DNA reference materials and datasets for comprehensively evaluating germline variant calling performance. Genome Biol. 24 , 270 (2023). Li, L. et al. The ChinaMAP reference panel for the accurate genotype imputation in Chinese populations. Cell Res 31 , 1308–1310 (2021). Li, J. et al. Limb development genes underlie variation in human fingerprint patterns. Cell 185 , 95–112.e18 (2022). Yang, J. et al. Conditional and joint multiple-SNP analysis of GWAS summary statistics identifies additional variants influencing complex traits. Nat. Genet. 44 , 369–75, S1–3 (2012). Sepp, M., Pruunsild, P. & Timmusk, T. Pitt-Hopkins syndrome-associated mutations in TCF4 lead to variable impairment of the transcription factor function ranging from hypomorphic to dominant-negative effects. Hum. Mol. Genet. 21 , 2873–2888 (2012). Lehalle, D., Williams, C., Siu, V. M. & Clayton-Smith, J. Fetal pads as a clue to the diagnosis of Pitt-Hopkins syndrome. Am. J. Med. Genet. A 155A , 1685–1689 (2011). Pitt, D. & Hopkins, I. A syndrome of mental retardation, wide mouth and intermittent overbreathing. Aust. Paediatr. J. 14 , 182–184 (1978). Sollis, E. et al. The NHGRI-EBI GWAS Catalog: knowledgebase and deposition resource. Nucleic Acids Res. 51 , D977–D985 (2023). Geoffroy, V. et al. AnnotSV: an integrated tool for structural variations annotation. Bioinformatics 34 , 3572–3574 (2018). Gray, M., Bliss, D. Z. & McNichol, L. Moisture-Associated Skin Damage: Expanding and Updating Practice Based on the Newest ICD-10-CM Codes. J. Wound Ostomy Continence Nurs. 49 , 143–151 (2022). Shmerling, M. et al. a New IFN-Inducible Human Member of the Ly6a Subfamily Expressed by Spleen Cells and Associated with Inflammation and Viral Resistance. Immunohorizons 6 , 253–272 (2022). Tu, Z. et al. Pro-inflammatory cytokine IL-6 regulates LMO4 expression in psoriatic keratinocytes via AKT/STAT3 pathway. Immun Inflamm Dis 11 , e1104 (2023). Boehncke, W.-H. & Schön, M. P. Psoriasis. Lancet 386 , 983–994 (2015). Halprin, K. M. Epidermal ‘turnover time’--a re-examination. Br. J. Dermatol. 86 , 14–19 (1972). Nestle, F. O., Kaplan, D. H. & Barker, J. Psoriasis. N. Engl. J. Med. 361 , 496–509 (2009). Fishilevich, S. et al. GeneHancer: genome-wide integration of enhancers and target genes in GeneCards. Database 2017 , (2017). Nishikori, S. et al. Resistance training rejuvenates aging skin by reducing circulating inflammatory factors and enhancing dermal extracellular matrices. Sci. Rep. 13 , 10214 (2023). Ichihashi, M. & Ando, H. The maximal cumulative solar UVB dose allowed to maintain healthy and young skin and prevent premature photoaging. Exp. Dermatol. 23 Suppl 1 , 43–46 (2014). Ly, B. C. K., Dyer, E. B., Feig, J. L., Chien, A. L. & Del Bino, S. Research Techniques Made Simple: Cutaneous Colorimetry: A Reliable Technique for Objective Skin Color Measurement. J. Invest. Dermatol. 140 , 3–12.e1 (2020). van Beek, N. et al. Thyroid hormones directly alter human hair follicle functions: anagen prolongation and stimulation of both hair matrix keratinocyte proliferation and hair pigmentation. J. Clin. Endocrinol. Metab. 93 , 4381–4388 (2008). Czarnywojtek, A. et al. The role of antithyroglobulin, antiperoxidase and anti-TSH receptor autoantibodies in amiodarone-induced thyrotoxicosis and amiodarone-induced hypothyroidism (A two-center study). Neuro Endocrinol. Lett. 36 , 677–681 (2015). Yan, X. et al. Peroxidasin is essential for eye development in the mouse. Hum. Mol. Genet. 23 , 5597–5614 (2014). Choi, Y., Chan, A. P., Kirkness, E., Telenti, A. & Schork, N. J. Comparison of phasing strategies for whole human genomes. PLoS Genet. 14 , e1007308 (2018). Browning, S. R. & Browning, B. L. Haplotype phasing: existing methods and new developments. Nat Rev Genet 12 , 703–714 (2011). Li, H. Aligning sequence reads, clone sequences and assembly contigs with BWA-MEM. arXiv [q-bio.GN] (2013) doi:10.48550/ARXIV.1303.3997. Sedlazeck, F. J. et al. Accurate detection of complex structural variations using single-molecule sequencing. Nat. Methods 15 , 461–468 (2018). Poplin, R. et al. Scaling accurate genetic variant discovery to tens of thousands of samples. bioRxiv (2017) doi:10.1101/201178. Van der Auwera, G. A. & O’Connor, B. D. Genomics in the Cloud: Using Docker, GATK, and WDL in Terra . (O’Reilly Media, 2020). Smolka, M. et al. Detection of mosaic and population-level structural variants with Sniffles2. Nat. Biotechnol. 42 , 1571–1580 (2024). Martin, M. et al. WhatsHap: fast and accurate read-based phasing. bioRxiv (2016) doi:10.1101/085050. Danecek, P. et al. Twelve years of SAMtools and BCFtools. Gigascience 10 , (2021). English, A. C., Menon, V. K., Gibbs, R. A., Metcalf, G. A. & Sedlazeck, F. J. Truvari: refined structural variant comparison preserves allelic diversity. Genome Biol. 23 , 271 (2022). Browning, B. L., Tian, X., Zhou, Y. & Browning, S. R. Fast two-stage phasing of large-scale sequence data. Am. J. Hum. Genet. 108 , 1880–1890 (2021). Purcell, S. et al. PLINK: a tool set for whole-genome association and population-based linkage analyses. Am. J. Hum. Genet. 81 , 559–575 (2007). Willer, C. J., Li, Y. & Abecasis, G. R. METAL: fast and efficient meta-analysis of genomewide association scans. Bioinformatics 26 , 2190–2191 (2010). Gong, J. et al. Long-read sequencing of 945 Han individuals identifies novel structural variants associated with phenotypic diversity and disease susceptibility. medRxiv (2024) doi:10.1101/2024.03.21.24304654. Kang, H. M. et al. Variance component model to account for sample structure in genome-wide association studies. Nat. Genet. 42 , 348–354 (2010). Ramos, E. et al. Replication of genome-wide association studies (GWAS) loci for fasting plasma glucose in African-Americans. Diabetologia 54 , 783–788 (2011). Lee, J. J. et al. Gene discovery and polygenic prediction from a genome-wide association study of educational attainment in 1.1 million individuals. Nat. Genet. 50 , 1112–1121 (2018). Cingolani, P. et al. A program for annotating and predicting the effects of single nucleotide polymorphisms, SnpEff: SNPs in the genome of Drosophila melanogaster strain w1118; iso-2; iso-3. Fly 6 , 80–92 (2012). Quinlan, A. R. & Hall, I. M. BEDTools: a flexible suite of utilities for comparing genomic features. Bioinformatics 26 , 841–842 (2010). Additional Declarations There is NO Competing Interest. Supplementary Files SupplementaryInformation.docx Supplementary Information SupplementaryData.xlsx Supplementary Data Cite Share Download PDF Status: Under Review Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6312842","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":440290049,"identity":"fee0fba0-80a7-4a3d-a544-7410fbaf51e8","order_by":0,"name":"Shaohua Fan","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA/0lEQVRIiWNgGAWjYDACCQYGZhDN3sDA+KACwgYJEqGF5wADs8EZUrWwSRClRX5288PHhW12DDzsZ49VHGyzljdnYD54m4fBLg+XFoM7x4yNZ7YlM/Dw5KXdONiWbrizgS3ZmochuRinFokEM2neNmYGe4Ycs9sf2w4zbjjAYybNw3AgsQGXw2akfwNqqWfg4X9jVnCw7bD9hgP83/BqYbiRA7LlMAOPRI4ZA1BLItAWNrxaDG7kFBvznDsO1PLGWOLAufTkDYfZjC3nGCTjc9jGxzxl1UCH5Rh+OFBmbbvhePPDG28q7HA7DArqEQrAUWNAQP0oGAWjYBSMArwAAE+DUk0hMur3AAAAAElFTkSuQmCC","orcid":"https://orcid.org/0000-0003-0610-9106","institution":"Fudan University","correspondingAuthor":true,"prefix":"","firstName":"Shaohua","middleName":"","lastName":"Fan","suffix":""},{"id":440290050,"identity":"c3d078fc-be13-4a9e-882b-8003820adb78","order_by":1,"name":"Tingting Gong","email":"","orcid":"https://orcid.org/0000-0001-5907-2445","institution":"Fudan University","correspondingAuthor":false,"prefix":"","firstName":"Tingting","middleName":"","lastName":"Gong","suffix":""},{"id":440290051,"identity":"d91b8ecc-6439-454c-b097-9842635b2d7b","order_by":2,"name":"Yulu Zhou","email":"","orcid":"","institution":"Fudan University","correspondingAuthor":false,"prefix":"","firstName":"Yulu","middleName":"","lastName":"Zhou","suffix":""},{"id":440290052,"identity":"53b43441-12d3-4837-b0b4-e0dddf1ac1d4","order_by":3,"name":"Junfan Zhao","email":"","orcid":"","institution":"Fudan University","correspondingAuthor":false,"prefix":"","firstName":"Junfan","middleName":"","lastName":"Zhao","suffix":""},{"id":440290053,"identity":"419bab72-8be5-4ad5-8a32-e8f6f16ed51d","order_by":4,"name":"Yechao Huang","email":"","orcid":"","institution":"Fudan University","correspondingAuthor":false,"prefix":"","firstName":"Yechao","middleName":"","lastName":"Huang","suffix":""},{"id":440290054,"identity":"e9fdb808-df7a-48e8-bfff-d81f04767a25","order_by":5,"name":"Jiao Gong","email":"","orcid":"","institution":"Fudan University","correspondingAuthor":false,"prefix":"","firstName":"Jiao","middleName":"","lastName":"Gong","suffix":""},{"id":440290055,"identity":"34b92130-eaf1-404b-9f0a-b9ff496b4650","order_by":6,"name":"Jinxi Li","email":"","orcid":"https://orcid.org/0000-0002-1366-9593","institution":"Fudan University","correspondingAuthor":false,"prefix":"","firstName":"Jinxi","middleName":"","lastName":"Li","suffix":""},{"id":440290056,"identity":"9a061c34-0d67-4e73-9ff8-0a84b73cd4c3","order_by":7,"name":"Qianqian Peng","email":"","orcid":"https://orcid.org/0000-0002-2018-5706","institution":"University of Chinese Academy of Sciences","correspondingAuthor":false,"prefix":"","firstName":"Qianqian","middleName":"","lastName":"Peng","suffix":""},{"id":440290057,"identity":"58c2a7b9-bd41-4af5-9974-487f7d059f63","order_by":8,"name":"Huidan Chang","email":"","orcid":"","institution":"Shanghai Institute of Nutrition and Health Chinese Academy of Sciences","correspondingAuthor":false,"prefix":"","firstName":"Huidan","middleName":"","lastName":"Chang","suffix":""},{"id":440290058,"identity":"353e90e3-3270-456c-8373-ecb29cfedb8f","order_by":9,"name":"Liyun Yuan","email":"","orcid":"","institution":"Shanghai Institute of Nutrition and Health Chinese Academy of Sciences","correspondingAuthor":false,"prefix":"","firstName":"Liyun","middleName":"","lastName":"Yuan","suffix":""},{"id":440290059,"identity":"104f5dc1-8e53-4209-8c1c-2aa8ec2bad6a","order_by":10,"name":"Guoqing Zhang","email":"","orcid":"","institution":"University of Chinese Academy of Sciences","correspondingAuthor":false,"prefix":"","firstName":"Guoqing","middleName":"","lastName":"Zhang","suffix":""},{"id":440290060,"identity":"749870da-7f94-40af-80e0-1ca05ddb6671","order_by":11,"name":"Sijia Wang","email":"","orcid":"https://orcid.org/0000-0001-6961-7867","institution":"University of Chinese Academy of Sciences","correspondingAuthor":false,"prefix":"","firstName":"Sijia","middleName":"","lastName":"Wang","suffix":""},{"id":440290061,"identity":"930601ee-a216-44fd-aca9-6a7083b96659","order_by":12,"name":"Li Jin","email":"","orcid":"https://orcid.org/0000-0001-9201-2321","institution":"Fudan University","correspondingAuthor":false,"prefix":"","firstName":"Li","middleName":"","lastName":"Jin","suffix":""}],"badges":[],"createdAt":"2025-03-26 13:35:23","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6312842/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6312842/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":80276734,"identity":"8506c6f0-cac3-446c-8dd2-3ac9e30052e2","added_by":"auto","created_at":"2025-04-10 05:06:37","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":117412,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eHybrid phasing workflow of Han-Chinese haplotype reference panel (Han-SV panel) construction. \u003c/strong\u003eWhole-genome SRS and LRS alignment data were sourced from a published Han Chinese cohort 19. The long-read-based co-phasing of small variants and SVs were implemented in individual-level and statistical-based phasing was applied in population-level. Dataset obtained in each step were shown in boxes with borders.\u003c/p\u003e","description":"","filename":"image1.png","url":"https://assets-eu.researchsquare.com/files/rs-6312842/v1/bbea4cf943f865006cc9b0a8.png"},{"id":80279010,"identity":"b20354ca-7e84-4c10-83b8-ba750d7b86d0","added_by":"auto","created_at":"2025-04-10 05:30:46","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":99516,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePhasing accuracy and variant count of small variants and structural variants in haplotype scaffold and Han-SV panel. \u003c/strong\u003e(A) Haplotype phasing accuracy, measured by switch error rate of small variants and flip error rate of SVs (DELs and INSs). (B) Haplotype phasing accuracy, measured by switch error rate of small variants within different MAF bins of HG002 (circle) and LCL6 (triangle) in long-read-based haplotype scaffold (blue), statistical-based haplotype set (green) and hybrid-based haplotype set (orange). The statistical phasing without long-read-based scaffold was for evaluation purposes only. Count of small variants (C) and SVs (D) in the Han-SV haplotype reference panel. Variant types were labelled in y-axis and their counts were in x-axis.\u003c/p\u003e","description":"","filename":"image2.png","url":"https://assets-eu.researchsquare.com/files/rs-6312842/v1/9a3b428cc80ef0ff0eef8dad.png"},{"id":80276738,"identity":"d013ae8e-44ee-4f77-9ec1-cd7d09005a6f","added_by":"auto","created_at":"2025-04-10 05:06:37","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":128500,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eImputation quality across different MAFs.\u003c/strong\u003e Count of imputed small variants (A) and SVs (B) on chromosome 20 of test 100 samples were shown across different MAF bins. Reference panels employed for imputation were labelled in y-axis and imputed counts were in y-axis. (C) Imputation accuracy (recall, precision and F1 score) of the high-quality (dosage r\u003csup\u003e2\u003c/sup\u003e ≥ 0.7) imputed variants on chromosome 20 of the test 100 samples, across different MAF bins. Reference panels employed for imputation were labelled in y-axis and imputed accuracy were in y-axis.\u003c/p\u003e","description":"","filename":"image3.png","url":"https://assets-eu.researchsquare.com/files/rs-6312842/v1/b2ba56edeac939d2d7aab0b6.png"},{"id":80276743,"identity":"25a93479-a6da-450b-a694-8b543363b0dc","added_by":"auto","created_at":"2025-04-10 05:06:37","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":796686,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eGWAS conducted using the variants obtained from the imputation of SNP array data. \u003c/strong\u003e\u0026nbsp;(A) LocusZoom plot of variants in the genomic region identified by the top variant rs17089876 associated with the left little finger (D5L) fingerprint. The lead variant rs17089876 is highlighted with a purple diamond, and the deletion (DEL) in linkage disequilibrium (LD) (r² = 0.68) with the lead variant is denoted by a triangle. Colors represent the strength of LD between rs17089876 and other variants. Genes in the region, including their intron-exon structure, transcription direction, and genomic coordinates (in Mb, NCBI human genome sequence, Build 38), are shown in the middle panel. (B) UCSC Genome Browser visualisation of the region. The browser displays the genomic location and transcript isoforms of \u003cem\u003eTCF4 \u003c/em\u003eacross the human reference assembly. Gene annotations include exons (boxes), introns (thin lines), and untranslated regions (intermediate-height boxes). The GeneHancer dataset is represented by grey boxes (enhancers) and red boxes (promoter regions). Yellow boxes at the bottom indicate clustered interactions between GeneHancer regulatory elements and genes. Red, yellow, and blue boxes at the bottom represent a subset of representative DNase hypersensitive sites across ENCODE data.\u003c/p\u003e","description":"","filename":"image4.png","url":"https://assets-eu.researchsquare.com/files/rs-6312842/v1/646d95b58e681c7953e1cb3d.png"},{"id":80279013,"identity":"d839562a-78f9-490d-aac0-2dd2b9a1e898","added_by":"auto","created_at":"2025-04-10 05:30:50","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":533554,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eGWAS conducted using the variants obtained from the imputation of SRS derived variants. \u003c/strong\u003e\u0026nbsp;(A) Left panel: LocusZoom plot of variants in the genomic region identified by the top 169 bp INS on chromosome 8 significantly associating (P-value \u0026lt; 1 × 10⁻\u003csup\u003e8\u003c/sup\u003e) with skin moisture. Right panel: Violin plot of the top 169 bp INS significantly associated with skin moisture. Each dot represents an individual. The 0/0, 0/1 and 1/1 genotypes were colored in red, green and blue, respectively. P-value was calculated using the EMMAX method. AU: arbitrary units, measure skin hydration in the stratum corneum, the smaller AU, the less water content of the stratum corneum.\u0026nbsp; (B) Left panel: LocusZoom plots variants in the genomic region identified by the top 103 bp DEL on chromosome 1 significantly associating (P-value \u0026lt; 1 × 10⁻\u003csup\u003e8\u003c/sup\u003e) with skin roughness in the flexor side of the left arm. Right panel: Violin plot of the top 103 bp DEL significantly associated with skin roughness. Each dot represents an individual. The 0/0, 0/1 and 1/1 genotypes were colored in red, green and blue, respectively. P-value was calculated using the EMMAX method. SEr: SELS-Roughness, measure skin roughness, the smaller SEr, the rougher the skin. (C) Left panel: LocusZoom plots for variants in the genomic region identified by the top 80 bp INS on chromosome 15 suggestively associating (P-value \u0026lt; 1 × 10⁻\u003csup\u003e5\u003c/sup\u003e) with skin smoothness in left check. Right panel: Violin plot of the top 80 bp INS suggestively associated with skin smoothness. The 0/0 and 0/1 genotypes were colored in red, green and blue, respectively. P-value was calculated using the EMMAX method. SEsm: SELS-Smoothness, measure skin smoothness, the smaller SEsm, the smoother the skin. (D) Left panel: LocusZoom plots for variants in the genomic region identified by the top 118 bp DEL on chromosome 2 suggestively associating (P-value \u0026lt; 1 × 10⁻\u003csup\u003e5\u003c/sup\u003e) with ITA° in the flexor side of the left arm. The GeneHancer dataset is represented by orange boxes (enhancers), red boxes (promoter regions) and blue boxes (enhancer or promoter regions) at the bottom. Right panel: Violin plot of the top 118 bp DEL suggestively associated with ITA°. The 0/0, 0/1 and 1/1 genotypes were colored in red, green and blue, respectively. P-value was calculated using the EMMAX method. LocusZoom plots: Top variant is highlighted with a purple diamond shape. Colors represent the strength of LD between each top variant and other variants. Genes in each region, their intron–exon structure, direction of transcription and genomic coordinates (in Mb, using the NCBI human genome sequence, Build 38, as reference) are shown in the middle.\u003c/p\u003e","description":"","filename":"image5.png","url":"https://assets-eu.researchsquare.com/files/rs-6312842/v1/e7a09fbc94ceed9b47bbf6d3.png"},{"id":80279170,"identity":"ea36d55c-28a1-4e46-9cb9-c3cf1c363b9e","added_by":"auto","created_at":"2025-04-10 05:33:04","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":3056926,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6312842/v1/ccc4f5b8-f63d-4e0e-8e63-937c50bbcef7.pdf"},{"id":80276737,"identity":"cd88a922-a646-406a-9f80-e4e30e02ebd5","added_by":"auto","created_at":"2025-04-10 05:06:37","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":824349,"visible":true,"origin":"","legend":"Supplementary Information","description":"","filename":"SupplementaryInformation.docx","url":"https://assets-eu.researchsquare.com/files/rs-6312842/v1/18c24ba8776a543a947e56fc.docx"},{"id":80276739,"identity":"ebddb338-0add-4a18-a099-c970cf94758e","added_by":"auto","created_at":"2025-04-10 05:06:37","extension":"xlsx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":178209,"visible":true,"origin":"","legend":"Supplementary Data","description":"","filename":"SupplementaryData.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-6312842/v1/66f5f8293a995d4c737c1efd.xlsx"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"A Long-read based Haplotype Panel Enhances Imputation and Discovery of Functional Small and Structural Variants","fulltext":[{"header":"Introduction","content":"\u003cp\u003eUnderstanding the genetic architecture and uncovering genotype-phenotype associations is one fundamental goal of human genetics. With the significant advancement in cataloguing of genetic variants among different populations and the extensive phenotypic characterization of cohorts, numerous genetic variants have been identified to be associated with human phenotypes through genome-wide association studies (GWAS). However, a considerable fraction of the genetic variability underling complex phenotypes still remains unexplored, as current GWAS primarily focus on single nucleotide variants (SNVs) and small insertions and deletions (InDels), often overlooking large structural variations (SVs), leading to critical gaps in missing heritability \u003csup\u003e1\u003c/sup\u003e. Notably, SVs have been reported to be 50 times more likely to affect gene expression than SNVs and three times more likely to be associated with phenotypic traits in GWAS \u003csup\u003e2,3\u003c/sup\u003e. Therefore, the comprehensive discovery and functional interpretation of SVs are crucial for addressing missing heritability in complex traits \u003csup\u003e4,5\u003c/sup\u003e and identifying key disease risk factors \u003csup\u003e6,7\u003c/sup\u003e. Incorporating SVs into GWAS is both feasible and essential to fully capture the spectrum of genetic variation and its contributions to complex traits.\u003c/p\u003e \u003cp\u003eIn population-level GWAS studies, imputation on lower-cost microarray or low-coverage sequencing data using haplotype reference panels has now been the most practical and powerful strategy to identify associated variants with phenotypes. As such, large-scale efforts combining improved sequencing methodologies are now including a much larger and richer spectrum of SV haplotypes in haplotype reference panels. In 2015, the 1000 genomes project (1KGP) was the first to include SVs in its haplotype panel, utilizing low-coverage (~\u0026thinsp;7.4X) whole-genome short-read sequencing (SRS) data of 2,504 individuals from 26 populations \u003csup\u003e1\u003c/sup\u003e. In 2022, 1KGP released an expanded reference panel, featuring high-coverage (30X) sequencing and including 602 trios \u003csup\u003e8\u003c/sup\u003e. In addition, great efforts have been made to develop population-specific reference panels using SRS data, such as Genome of the Netherlands (GoNL), Genome for Life Cohort (GCAT) and BioBank Japan (BBJ) developing SV-enriched haplotype reference panel for Dutch, Iberian and Japanese population, respectively \u003csup\u003e9\u0026ndash;11\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eHowever, these studies primarily rely on SRS technology, which can lead to an underrepresentation of SVs in reference panels. This is due to the short read length and the limitation of SRS to resolve highly repetitive and complex genomic regions \u003csup\u003e12,13\u003c/sup\u003e. Moreover, the large size of SVs\u0026mdash;often exceeding 1 Mbp\u0026mdash;contrasts with the relatively small haplotype blocks (typically 1\u0026ndash;10 kb) generated through statistical phasing \u003csup\u003e14\u003c/sup\u003e, resulting in only a limited number of SV haplotypes being inferred from SRS data, often with suboptimal accuracy \u003csup\u003e15,16\u003c/sup\u003e. Although some haplotype panels, such as the expanded 1KGP and GoNL-SV panels, incorporate parental data to enhance phasing accuracy, SV haplotypes still rely heavily on statistical inference \u003csup\u003e8,9\u003c/sup\u003e. These limitations in SV detection and phasing ultimately reduce the imputation performance of haplotype panels, weakening the power of GWAS and fine-mapping causal variants. This highlights the need for haplotype panels that comprehensively capture the full spectrum of SV haplotypes in the human genome.\u003c/p\u003e \u003cp\u003eLong-read sequencing technologies (LRS), capable of sequencing DNA fragments over 10 kb in length, far exceed the read lengths achievable with SRS, which are typically around 150 bp. LRS has been shown to significantly improve SV discovery \u003csup\u003e17\u0026ndash;19\u003c/sup\u003e and provide access to genomic regions that are challenging to analyze with SRS data \u003csup\u003e12\u003c/sup\u003e. The comprehensive SV characterization based on LRS has offered new insights into SV functional interpretation, such as a rare DEL in \u003cem\u003ePCSK9\u003c/em\u003e found associated with cholesterol level \u003csup\u003e17\u003c/sup\u003e and causal SVs impacting \u003cem\u003eGSDMD\u003c/em\u003e for bone density and \u003cem\u003eWWP2\u003c/em\u003e for height, weight, fat, craniofacial phenotypes, and innate immunity \u003csup\u003e19\u003c/sup\u003e. In addition, LRS has demonstrated its ability to significantly enhance haplotype phasing accuracy via direct read-based phasing without parents or large cohort data \u003csup\u003e20,21\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eBuilding upon our previous work, which utilized whole-genome Oxford Nanopore long-read sequencing to generate a comprehensive SV set from 945 Han Chinese genomes \u003csup\u003e19\u003c/sup\u003e, this study provides a haplotype reference panel (Han-SV panel) to facilitate variant imputation for Chinese populations. Through hybrid phasing strategy combining long-read-based and statistical-based phasing methods, we achieve high phasing accuracy for both small variants and SVs in this Han-SV panel. Imputation performance evaluation of the Han-SV panel demonstrates its potential to significantly enhance SV imputation quality. We further conducted two SV-included GWASs, showing the value of our Han-SV panel to improve the power of GWAS and facilitate the identification of previously unreported variants associated with complex phenotypes and diseases.\u003c/p\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eHaplotype panel construction and evaluation\u003c/h2\u003e \u003cp\u003eThis study included 943 individuals with both whole genome SRS and LRS data from our previously published Han Chinese cohort \u003csup\u003e19\u003c/sup\u003e. From SRS data (~\u0026thinsp;30X coverage), we identified 37,536,312 high-quality small variants, comprising 34,167,444 SNVs and 3,368,868 InDels, with a median of 3,987,826 SNVs and 931,557 InDels per sample (see \u003cb\u003eMethods, Supplementary Fig.\u0026nbsp;1\u003c/b\u003e). Additionally, using Oxford Nanopore long-read sequencing data (100 individuals with ~\u0026thinsp;30X and 843 with ~\u0026thinsp;15X coverage), we identified 19,829 SVs per individual, with a median of 8,167 DELs, 10,923 INSs, 588 DUPs, and 155 INVs after quality control (see \u003cb\u003eMethods\u003c/b\u003e, \u003cb\u003eSupplementary Fig.\u0026nbsp;1\u003c/b\u003e).\u003c/p\u003e \u003cp\u003eTo generate a haplotype panel for the small variant and SV call sets, we applied a hybrid phasing workflow integrating long-read-based and statistical-based phasing methods (see Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e and \u003cb\u003eMethods\u003c/b\u003e). We first performed long-read-based co-phasing of SNVs, InDels and SVs using LongPhase \u003csup\u003e21\u003c/sup\u003e individually and merged the long-read phased haplotypes into one population-level haplotype set. Next, we applied SHAPEIT5 \u003csup\u003e22\u003c/sup\u003e to statistically infer haplotypes for unphased variants, using the long-read phased variants as haplotype scaffold.\u003c/p\u003e \u003cp\u003eWe also included samples of HG002 from Genome in a Bottle (GIAB) Consortium \u003csup\u003e23,24\u003c/sup\u003e and LCL6 from the Chinese Quartet Genome Project \u003csup\u003e25\u003c/sup\u003e in our phasing workflow, with the purpose to access the phasing accuracy using their developed benchmark haplotype call sets. The GIAB developed HG002 benchmark haplotype-resolved set includes 2,869,654 bi-allelic heterozygous variants (2,360,987 SNVs, 492,454 InDels and 16,213 SVs) (see \u003cb\u003eSupplementary Data 1\u003c/b\u003e). The haplotype-resolved variant set of LCL6 was generated from trio-based phasing with parents\u0026rsquo; data (see \u003cb\u003eMethods)\u003c/b\u003e, resulting in a total of 2,035,610 (2,029,241 SNVs and 6,369 SVs) bi-allelic heterozygous variants (see \u003cb\u003eSupplementary Data 1\u003c/b\u003e).\u003c/p\u003e \u003cp\u003eTo access the variant phasing accuracy, published HG002 and LCL6 small variant sets were integrated into our SRS-called small variant set, which then was processed through the same phasing workflow. While raw ONT sequencing reads of HG002 and LCL6 were used for the same read alignment and SV calling, genotyping and phasing workflow as all samples in our Han Chinese cohort. The phasing accuracies were then assessed by comparing the phased set of HG002 and LCL6 to their benchmark haplotype sets as the reference truth set (see \u003cb\u003eMethods\u003c/b\u003e).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eIndividual-level phasing and haplotype scaffold construction with long reads\u003c/h3\u003e\n\u003cp\u003eThe generated long-read-based haplotype set included 36,265,725 small variants and 204,034 SVs in total, in which 21,515,772 (59.3%) small variants and 3,961 (1.9%) SVs were phased for all samples. When examining the phasing rates individually, we observed an average of 87% small variants (range 80% \u0026minus;\u0026thinsp;92%) and 84% SVs (range 64% \u0026minus;\u0026thinsp;87%) per-individual were phased using long-read-based phasing (see \u003cb\u003eSupplementary Data 2\u003c/b\u003e). Among different types in SVs, DELs (92%, range 66\u0026ndash;94%) and INSs (79%, range 62\u0026ndash;83%) had higher average phasing rates compared to DUPs (60%, range 36\u0026ndash;75%) and INV (67%, range 44\u0026ndash;88%) (see \u003cb\u003eSupplementary Data 2\u003c/b\u003e).\u003c/p\u003e \u003cp\u003eWe assessed the phasing accuracy of the small variant haplotype scaffold by calculating the switch error rate (SER) for HG002 and LCL6, using their benchmark haplotype sets as the reference truth set. In HG002, read-based co-phasing of SNVs, InDels and SVs using LongPhase achieved an average SER of 0.09% for small variants, while LCL6 showed a slightly higher average SER of 0.14% (see Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eA). We observed that multi-type variants co-phasing using LongPhase have lower SERs than read-based SNV-only phasing using WhatsHap (0.12% for HG002 and 0.21% for LCL6). Compared to previous SRS-based study of expanded-1KGP \u003csup\u003e8\u003c/sup\u003e (see Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e), the SERs of small variants phasing in this study are close to the SERs of children (average 0.09%) from trio-based phasing, but notably lower than the reported SERs of unrelated individuals (average 0.79%) from statistical phasing.\u003c/p\u003e \u003cp\u003e \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\u003eComparison of variant phasing accuracy with different methods\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" 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 \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePhasing method\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSmall variant SER\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eDEL flip rate\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eINS flip rate\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLong-read-based multi-variant types co-phasing by LongPhase\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.09\u0026ndash;0.14%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.22\u0026ndash;0.38%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.34\u0026ndash;0.69%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLong-read-based phasing by WhatsHap\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.12\u0026ndash;0.21%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eNA\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHybrid phasing (statistical and long-read-based)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.32\u0026ndash;0.33%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.36\u0026ndash;0.52%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.49\u0026ndash;2.42%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eStatistical phasing on unrelated individuals\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.65\u0026ndash;1.66%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.80\u0026ndash;4.75%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1.18\u0026ndash;6.01%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTrio-based phasing\u003c/p\u003e \u003cp\u003e(Expanded-1KGP)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.07\u0026ndash;0.09%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eNA\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHybrid phasing\u003c/p\u003e \u003cp\u003e(statistical and trio-based, Expanded-1KGP)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.79%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.89%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.24%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eStatistical phasing on unrelated individuals\u003c/p\u003e \u003cp\u003e(BBJ-SV panel)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.7%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e3.5%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eThe phasing accuracy of SVs in the haplotype scaffold was evaluated by calculating the flip error rate relative to the benchmark SV haplotype set of HG002 and LCL6. Comparing SVs in the haplotype scaffold and benchmark haplotype set, we defined two SVs as concordant if they are in the same type and breakpoints difference\u0026thinsp;\u0026lt;\u0026thinsp;10 bp, accounting for breakpoint shift in SV reporting mainly due to micro-homology around SV breakpoint. We observed that 19% of the SVs (1,511 DELs and 1,577 INSs) in HG002 and 40% of the SVs (1,348 DELs and 1,135 INSs) in LCL6 in the benchmark sets as concordant in our haplotype scaffold (see \u003cb\u003eSupplementary Fig.\u0026nbsp;2\u003c/b\u003e). Our analysis indicates a flip rate of 0.22% for DELs and 0.69% for INSs in HG002. For LCL6, we estimated flip rates of 0.38% and 0.34% for DELs and INSs, respectively (see Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eA). We found that long-read-based phasing of DELs had lower flip rates, while INSs exhibited greater flip rates than the rate reported by the expanded 1KGP (0.89% of DELs and 0.24% of INSs), where SV haplotypes were inferred with statistical phasing using a trio-based small variant haplotype scaffold (see Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). Additionally, our approach yielded lower flip rates for both DELs and INSs when compared to statistical phasing with unrelated samples (flip rate\u0026thinsp;=\u0026thinsp;1.7% for DELs and 3.5% for INSs) as reported by BBJ-SV study \u003csup\u003e11\u003c/sup\u003e (see Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e\n\u003ch3\u003ePopulation-level haplotype phasing and panel development with hybrid approach\u003c/h3\u003e\n\u003cp\u003eWith this hybrid phasing approach, the resulting integrated haplotype panel (Han-SV panel), consisting of 35,783,527 small variants and 172,569 SVs after quality control \u003cb\u003e(\u003c/b\u003esee Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eC and \u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eD\u003cb\u003e).\u003c/b\u003e With the significant increase in the number of SVs per individual identified using LRS data in this study, we achieved nearly 5-fold more phased heterozygous SVs per individual in the Han-SV panel (16,998) compared to the expanded 1000 Genomes Project panel (3,462). The total number of SVs in our Han-SV panel is nearly two-fold more than that in the expanded-1KGP panel (97,862 across autosomes).\u003c/p\u003e \u003cp\u003eThe phasing accuracy of small variants resulting from the hybrid phasing approach was again evaluated based on the SERs of HG002 and LCL6. The SERs of small variants increased after statistical phasing (SER\u0026thinsp;=\u0026thinsp;0.33% HG002 and SER\u0026thinsp;=\u0026thinsp;0.32% for LCL6) compared to read-based phasing alone (see Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eA). Examining the SERs across different MAF bins (see Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eB), we found rare (MAF\u0026thinsp;\u0026lt;\u0026thinsp;1%) small variants have lower phasing accuracy (1.5% for HG002 and 8.3% for LCL6), comparing to low-frequency (1% \u0026le; MAF\u0026thinsp;\u0026lt;\u0026thinsp;5%) variants (0.78% for HG002 and 3.6% for LCL6) and common (MAF\u0026thinsp;\u0026ge;\u0026thinsp;5%) ones (0.23% and 0.26%). When compared against statistical phasing without the long-read-based scaffold, this hybrid phasing approach improved phasing accuracy by around 2-fold for LCL6 and 5-fold for HG002 on average and across different MAF bins (see Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eA and \u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eB). The improvement in phasing accuracy using this hybrid phasing approach is also confirmed by comparing to the reported SERs by the expanded-1KGP (see Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e), showing 2.4-fold decrease in SERs of small variants.\u003c/p\u003e \u003cp\u003eWe further evaluated the SV phasing accuracy by the hybrid phasing using flip error rate of HG002 and LCL6 (see \u003cb\u003eMethods\u003c/b\u003e) \u003csup\u003e8\u003c/sup\u003e. Overall, 19% of the SVs (1,384 DELs and 1,681 INSs) in HG002 and 25% of the SVs (571 DELs and 1,031 INSs) in LCL6 in the benchmark sets identified as concordant in our haplotype set and were compared to assess the flip error rate (see \u003cb\u003eSupplementary Fig.\u0026nbsp;2\u003c/b\u003e). We observed that DEL phasing flip error rates increased by around 0.3% on average in HG002 and remained similar in LCL6 with additional statistical phasing, while INS flip error rates increased by 1.73% in HG002 and increased by 0.16% in LCL6, compared to long-read-based phasing alone (see Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eA). When compared against statistical phasing without the long-read-based scaffold, this hybrid phasing approach improved SV phasing accuracy around 5 to 9-fold for DELs and 2.4 to 5-fold for INSs (see Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eA and Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). Compared to published SV-included haplotype panels (see Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e), we achieved lower flip rates of DEL compared to the expanded-1KGP study and for both DELs and INSs compared to BBJ-SV study \u003csup\u003e11\u003c/sup\u003e.\u003c/p\u003e\n\u003ch3\u003eImputation performance evaluation of the haplotype panel\u003c/h3\u003e\n\u003cp\u003eTo evaluate the imputation performance of our Han-SV panel, we utilized variants from 100 samples of this cohort on chromosome 20 with available WGS data as the imputation target. Variants from the remaining 843 samples were used to construct a test Han-SV panel using the hybrid phasing workflow (see \u003cb\u003eMethods\u003c/b\u003e and \u003cb\u003eSupplementary Fig.\u0026nbsp;3\u003c/b\u003e). Another two panels were employed for comparison: the expanded-1KGP panel and ChinaMAP. The expanded-1KGP is a widely used multi-ancestry reference panel, which includes high-quality SV haplotypes by high coverage SRS (30X) and trio data. While ChinaMAP \u003csup\u003e26\u003c/sup\u003e is one SNV imputation panel for the Chinese population, constructed with high coverage SRS data (41X).\u003c/p\u003e \u003cp\u003eOverall, our panel imputed fewer total SNVs than expanded-1KGP and ChinaMAP (see Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eA \u003cb\u003eand Supplementary Fig.\u0026nbsp;4\u003c/b\u003e). Assessing their imputation quality, defined by dosage r\u0026sup2;, revealed that our panel imputed a similar number of SNVs with dosage r\u0026sup2; \u0026ge; 0.1 to 0.9 (see \u003cb\u003eSupplementary Fig.\u0026nbsp;4\u003c/b\u003e). Defining high-quality imputed variants as dosage r\u0026sup2; \u0026ge; 0.7, our panel imputed 163,565 SNVs, which is 12% more than the proportion of high-quality SNVs than the expanded-1KGP (159,611) and ChinaMAP (157,128). For InDels, although fewer high-quality InDels (15,637 vs. 45,799) were imputed compared to the expanded-1KGP panel, our panel achieved a comparable high-quality rate (24% vs 22%). In addition, our panel imputed 1.75-fold more total SVs and 4.3-fold more high-quality SVs, compared to the expanded-1KGP panel (see Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eB \u003cb\u003eand Supplementary Fig.\u0026nbsp;4\u003c/b\u003e). As ChinaMAP is a SNV imputation panel, it has no InDels or SVs imputed (see Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eImputation quality with different reference haplotype panels\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" 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 \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVariant Type\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eImputation Count and Quality\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eExpanded-1KGP\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eChinaMAP\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eTest Han-SV panel\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e \u003cp\u003e\u003cb\u003eSNV\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTotal Count\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1,365,653\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1,353,675\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e690,259\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eHigh-quality (dosage r\u0026sup2; \u0026ge; 0.7) Count\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e159,611\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e157,128\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e163,565\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eF1 score\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e74%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e72%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e76%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e \u003cp\u003e\u003cb\u003eInDel\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTotal Count\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e205,380\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e65,379\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eHigh-quality (dosage r\u0026sup2; \u0026ge; 0.7) Count\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e45,799\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e15,637\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eF1 score\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e40%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eNA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e74%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e \u003cp\u003e\u003cb\u003eSV\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTotal Count\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2,203\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e3,874\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eHigh-quality (dosage r\u0026sup2; \u0026ge; 0.7) Count\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e292\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1,265\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eF1 score\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e12%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eNA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e53%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eTo assess the accuracy of variant imputation, we compared the high-quality (dosage r\u0026sup2; \u0026ge; 0.7) imputed variants from each panel to their corresponding WGS-derived variants, and evaluated recall, precision, and F1 score across for each sample (see \u003cb\u003eMethods\u003c/b\u003e, Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eC). Overall, our panel achieved a similar level of accuracy (72% \u0026minus;\u0026thinsp;76%), measured by F1 score, in SNV imputation, compared to the expanded-1KGP panel and ChinaMAP (see Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). As ChinaMAP is not able to impute InDels or SVs, we compared them to the expanded-1KGP panel (see Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e) and observed a total of 34% increase of accuracy in InDel imputation and 41% increase in SV imputation by our panel.\u003c/p\u003e \u003cp\u003eWhen examining different MAF bins, Han-SV and other two panels had higher imputation accuracy for common variants (MAF\u0026thinsp;\u0026ge;\u0026thinsp;5%), while lower for low-frequency (1% \u0026le; MAF\u0026thinsp;\u0026lt;\u0026thinsp;5%) and rare (MAF\u0026thinsp;\u0026lt;\u0026thinsp;1%) variants (see Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eC). Specifically, all three panels achieved nearly 80% imputation accuracy (F1 score) for common SNVs, while around 50% for low-frequency and only 10% for rare SNVs. For InDels, Han-SV panel had similar imputation accuracy as SNVs in different MAF bins, while the expanded-1KGP panel had lower F1 scores with 41% accuracy for common, 31% for low-frequency and 9.0% for rare InDel imputation. The accuracy of SV imputation is lower than other variant types in each MAF bins, with 54% for common, 25% for low-frequency and 12% for rare SVs by Han-SV panel. It should be noted our Han-SV panel imputed 4.3 times more high-quality SVs than the expanded 1KGP panel (see Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eB), therefore, leading to a significantly higher overall sensitivity (37.6% vs. 6.7%) (see Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eC). With the higher sensitivity in SV imputation, our panel is able to maintain the similar level of high precision for common SVs (\u0026gt;\u0026thinsp;85%) and low-frequency SVs (around 60%) as the expanded 1KGP panel. Moreover, a 13% higher precision (24% vs 11%) for rare SVs, compared to the expanded 1KGP panel can be achieved by our panel.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e\n\u003ch3\u003eStructural variant imputation and GWAS application on SNP array data\u003c/h3\u003e\n\u003cp\u003eTo demonstrate the use and assess the value of Han-SV haplotype panel as an imputation resource for GWAS, we first performed variant imputation on SNP array data from a previously published GWAS study of fingerprint patterns (arch, loop, and whorl) of 10 fingers \u003csup\u003e27\u003c/sup\u003e based on 9,909 individuals from three Han Chinese cohorts (see \u003cb\u003eMethods\u003c/b\u003e). Overall, our panel imputed a comparable average number of SNVs (8,757,554 vs. 9,137,578) and fewer InDels (867,549 vs. 1,901,440) with dosage r\u0026sup2; \u0026ge; 0.7 in each cohort, while achieving approximately 2.3-fold more SVs (30,046 vs. 13,109) compared to the expanded-1000 Genomes Project (1KGP) panel (see \u003cb\u003eSupplementary Data 3\u003c/b\u003e).\u003c/p\u003e \u003cp\u003eWe used the same significance threshold of GWAS (P-meta\u0026thinsp;\u0026lt;\u0026thinsp;1.67 \u0026times; 10⁻⁸) and compared our results with those from the original publication (see \u003cb\u003eSupplementary Data 4, Methods\u003c/b\u003e) \u003csup\u003e27\u003c/sup\u003e. A total of two SVs and 707 small variants (643 SNPs and 64 InDels) passed the significance threshold. Additionally, stepwise selection analysis using GCTA-COJO (v1.94.1) \u003csup\u003e28\u003c/sup\u003e identified 17 independent variants, which are all SNPs. Of the 18 previously reported fingerprint-associated loci \u003csup\u003e27\u003c/sup\u003e, we successfully replicated 15 of them. The remaining three loci were not replicated in our study due to low imputation qualities of the variants in these regions (dosage r\u0026sup2; \u0026lt; 0.7) using Han-SV panel and were thus excluded from the GWAS analysis. Additionally, we identified two previously unreported association regions with lead SNPs achieving genome-wide significance (P-meta\u0026thinsp;\u0026lt;\u0026thinsp;1.67 \u0026times; 10⁻⁸) (see \u003cb\u003eSupplementary Data 4\u003c/b\u003e).\u003c/p\u003e \u003cp\u003eAlthough no SV reached genome-wide significant level (P-meta\u0026thinsp;\u0026lt;\u0026thinsp;1.67 \u0026times; 10⁻⁸) in the two previously unreported loci, we identified a 52 bp DEL on chromosome 18 (chr18: 55,586,143\u0026thinsp;\u0026minus;\u0026thinsp;55,586,194, GRCh38) within a previously identified locus that is marginally significantly associated with the fingerprint phenotype of the left little finger (D5L) (P-meta\u0026thinsp;=\u0026thinsp;3.66 \u0026times; 10⁻\u003csup\u003e8\u003c/sup\u003e) (see Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eA). The DEL is in linkage disequilibrium (r\u0026sup2; = 0.68) with the lead variant, rs17089876 (P-meta\u0026thinsp;=\u0026thinsp;1.53 \u0026times; 10⁻\u003csup\u003e9\u003c/sup\u003e) in the original study \u003csup\u003e27\u003c/sup\u003e. The DEL is located in an enhancer region of \u003cem\u003eTCF4\u003c/em\u003e based on the annotation of the GeneHancer database, potentially disrupting its expression (see Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eB). Previous studies have shown that pathogenic mutations in \u003cem\u003eTCF4\u003c/em\u003e cause Pitt-Hopkins syndrome, which is characterized by persistent fetal digital pads \u003csup\u003e29\u0026ndash;31\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eStructural variant imputation and GWAS application on SRS data\u003c/h2\u003e \u003cp\u003eWe also performed variant imputation on a SRS data called variant set, including 14,699,131 SNPs of an independent cohort with 1,016 samples of Han ancestry (see \u003cb\u003eMethods\u003c/b\u003e). A total of 70,949 SVs, 1,544,670 InDels and additional 7,536,153 SNPs were imputed with high-quality (dosage r\u0026sup2; \u0026ge; 0.7) (see \u003cb\u003eSupplementary Data 5\u003c/b\u003e).\u003c/p\u003e \u003cp\u003eWe selected 145 continuous skin phenotypes and conducted a GWAS for each skin trait using both SRS-derived and imputed variants. Only high-quality imputed and SRS-called variants with MAF\u0026thinsp;\u0026ge;\u0026thinsp;1% (48,503 SVs and 8,158,252 SNPs/InDels) were included. As a result, three SVs and 669 small variants (617 SNPs and 52 InDels) reached genome-wide significance (P-value\u0026thinsp;\u0026lt;\u0026thinsp;5 \u0026times; 10⁻⁸) across 73 traits (see \u003cb\u003eMethods\u003c/b\u003e). We further performed stepwise selection analysis using GCTA-COJO (v1.94.1) \u003csup\u003e28\u003c/sup\u003e, identifying 219 independent variants (2 SVs and 217 SNPs/InDels) across 73 traits (see \u003cb\u003eSupplementary Data 6\u003c/b\u003e). Defining association regions as \u0026plusmn;\u0026thinsp;250kb around the independent variants (P-value\u0026thinsp;\u0026lt;\u0026thinsp;5 \u0026times; 10⁻⁸), we found that 104 regions (51.2%) in the present study overlap with the reported variants associated with skin phenotypes in previous GWAS \u003csup\u003e32\u003c/sup\u003e, while 99 regions (48.8%) were unreported.\u003c/p\u003e \u003cp\u003eWe found two association regions were identified by independent SVs using GCTA-COJO (v1.94.1) \u003csup\u003e28\u003c/sup\u003e \u0026mdash;one previously reported and one unreported. Specifically, within a reported region, a 169 bp INS at chr8:143,072,208 (GRCh38) was found associated with the skin moisture phenotype (β = -3.381, P-value\u0026thinsp;=\u0026thinsp;2.16 \u0026times; 10⁻⁸) \u003cb\u003e(\u003c/b\u003esee Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eA \u003cb\u003eand Supplementary Fig.\u0026nbsp;5A)\u003c/b\u003e. Functional annotation using AnnotSV \u003csup\u003e33\u003c/sup\u003e suggested that the 169 bp insertion causes a frameshift in two isoforms of \u003cem\u003eLY6S\u003c/em\u003e (XM_054328425.1 and XM_054328424.1). The direct role of \u003cem\u003eLY6S\u003c/em\u003e in skin moisture remains unclear. However, LY6 family proteins typically participate in cellular processes, including immune responses, which may indirectly affect skin health and moisture regulation \u003csup\u003e34,35\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eIn addition, we identified a 103 bp DEL (chr1:87,581,439\u0026thinsp;\u0026minus;\u0026thinsp;87,581,541, GRCh38), is significantly (β = -0.538, P-value\u0026thinsp;=\u0026thinsp;3.15 \u0026times; 10⁻\u003csup\u003e10\u003c/sup\u003e) associated with skin roughness in an unreported region \u003cb\u003e(\u003c/b\u003esee Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eB \u003cb\u003eand Supplementary Fig.\u0026nbsp;5B)\u003c/b\u003e. The DEL is located in an intergenic region, and the nearest gene is \u003cem\u003eLMO4\u003c/em\u003e, as annotated by AnnotSV \u003csup\u003e33\u003c/sup\u003e. Previous studies have shown that \u003cem\u003eLMO4\u003c/em\u003e regulates keratinocyte proliferation and differentiation. Additionally, an overexpression of \u003cem\u003eLMO4\u003c/em\u003e has been observed in patients with psoriasis, a condition characterized by skin roughness \u003csup\u003e36\u0026ndash;39\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eGiven our relatively small cohort (1,016 samples), we applied a more lenient significance threshold of 1 \u0026times; 10⁻\u003csup\u003e5\u003c/sup\u003e, uncovering an additional 64 SVs, 11,382 SNPs, and 1,769 InDels, resulting in a total of 3,951 independent variants (37 SVs and 3,914 SNPs/InDels) by GCTA-COJO (v1.94.1) \u003csup\u003e28\u003c/sup\u003e. Defining association regions as \u0026plusmn;\u0026thinsp;250kb around independent variants, we observed 988 (55%) previously reported regions and 808 (45%) unreported regions compared to the reported skin related studies in NHGRI-EBI GWAS catalogue \u003csup\u003e28,32\u003c/sup\u003e (see \u003cb\u003eMethods\u003c/b\u003e). Overall, 36 association regions were identified through 37 independent SV signals, including 23 previously reported and 13 unreported regions (see \u003cb\u003eSupplementary Data 7\u003c/b\u003e).\u003c/p\u003e \u003cp\u003eOf the 35 independent signals identified with P-values ranging from 1 \u0026times; 10⁻\u003csup\u003e5\u003c/sup\u003e to 5 \u0026times; 10⁻\u003csup\u003e8\u003c/sup\u003e, functional annotation using AnnotSV revealed that nine SVs disrupt exons or enhancers of genes \u003csup\u003e33,40\u003c/sup\u003e (see \u003cb\u003eSupplementary Data 7, Methods\u003c/b\u003e). For example, we detected one 80 bp INS (chr15:101,185,413\u0026thinsp;\u0026minus;\u0026thinsp;101,185,413, GRCh38) associated with skin smoothness (β = -43.524, P-value\u0026thinsp;=\u0026thinsp;6.85 \u0026times; 10⁻\u003csup\u003e6\u003c/sup\u003e) which was located in the first exon of \u003cem\u003eCHSY1\u003c/em\u003e (XM_024449873.1), causing a frameshift mutation (see Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eC \u003cb\u003eand Supplementary Fig.\u0026nbsp;5C\u003c/b\u003e). Previous studies have shown that low expression of \u003cem\u003eCHSY1\u003c/em\u003e is associated with decreased skin thickness in aged skin \u003csup\u003e41\u003c/sup\u003e. Aged skin is characterized by the presence of mottled pigmentation, solar lentigines, rhytides (wrinkles), and xerosis (dryness), in contrast to the smoother, more even-toned appearance of younger skin. \u003csup\u003e41,42\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eIn addition, a 118 bp DEL (chr2:1,511,152-1,511,269, GRCh38) was positively associated with skin pigmentation based on Individual Typology Angle, ITA\u0026deg; (β\u0026thinsp;=\u0026thinsp;1.451, P-value\u0026thinsp;=\u0026thinsp;4.72 \u0026times; 10⁻\u003csup\u003e6\u003c/sup\u003e) (see Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eD \u003cb\u003eand Supplementary Fig.\u0026nbsp;5D\u003c/b\u003e). Higher ITA\u0026deg; values correspond to lighter skin, while lower values correspond to darker skin \u003csup\u003e43\u003c/sup\u003e. The DEL is located in the intron 16 of \u003cem\u003eTPO\u003c/em\u003e, overlapping an enhancer (GH02J001509) annotated by GeneHancer database that potentially regulates \u003cem\u003eTPO\u003c/em\u003e and \u003cem\u003ePXDN\u003c/em\u003e expression according to the Genehancer database \u003csup\u003e40\u003c/sup\u003e. Previous studies suggest that \u003cem\u003eTPO\u003c/em\u003e is involved in thyroid hormone production, which influences hair follicle pigmentation \u003csup\u003e44,45\u003c/sup\u003e, and \u003cem\u003ePXDN\u003c/em\u003e mutations can result in the white spot at the belly in mice \u003csup\u003e46\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eIn this study, we developed an SV-enriched haplotype reference panel for the Han Chinese population (Han-SV panel). Leveraging LRS and SRS data from 943 individuals, we combined long-read-based approaches with statistical phasing to accurately phase both small variants and SVs. This comprehensive haplotype set, alongside our integrated and highly accurate phasing method, represents a valuable resource for precise variant imputation and supports the inclusion of SVs in future GWAS.\u003c/p\u003e \u003cp\u003eBased on Oxford Nanopore long-read sequencing data, we firstly demonstrated that direct long-read-based phasing of small variants has a similar high level of phasing accuracy (SERs\u0026thinsp;=\u0026thinsp;0.09\u0026ndash;0.14%) as SRS-read-based phasing (0.10%) \u003csup\u003e47\u003c/sup\u003e and trio-based phasing (0.07\u0026ndash;0.09%) \u003csup\u003e8\u003c/sup\u003e. Without the extra cost on parents\u0026rsquo; data collection and sequencing, read-based phasing is further more feasible than trio-based phasing to be applied in population-scale sequencing projects\u003csup\u003e48\u003c/sup\u003e. More importantly, long-read-based phasing facilitates the co-phasing of SVs with small variants, which is difficult with short-read-based phasing \u003csup\u003e14,21\u003c/sup\u003e. Using long-read-based co-phasing of SNVs, InDels and SVs in this study, we showed that the error rate of long-read-based SV phasing (0.22% \u0026minus;\u0026thinsp;0.69%) was lower than statistical-based phasing of unrelated individuals (1.7% \u0026minus;\u0026thinsp;3.5%) \u003csup\u003e11\u003c/sup\u003e and close to phasing with trio data (0.24% \u0026minus;\u0026thinsp;0.89%) \u003csup\u003e8\u003c/sup\u003e. The use of long-read sequencing data to improve SV phasing have also been shown by other studies, using long-read data only on patients\u003csup\u003e20\u003c/sup\u003e, integrating long reads with Hi-C data and/or single-strand genome sequencing data\u003csup\u003e3,13\u003c/sup\u003e. However, these methods have only been applied in individual-level or small cohorts with less than 50 individuals.\u003c/p\u003e \u003cp\u003eIn this study, we observed that approximately 90% of small variants and 80% of SVs can be phased based on direct long-read sequencing reads at the individual level. However, when variants from multiple individuals are combined into a single multi-sample variant set, only 60% of small variants and 1.9% of SVs are phased across all individuals. The lower population-level phasing rate of SVs may result from variations in long-read data alignment quality at the same SV site across different samples, as well as individual differences in breakpoints, both of which contribute to reduced phasing rates when merging SVs from multiple individuals. To facilitate population-level phasing and haplotype panel construction, we employed a hybrid phasing strategy that integrates both read-based and statistical approaches, by building a long-read-based scaffold for statistical phasing on unrelated individuals. Compared to the read-based method alone, this significantly improved the phasing rate at the population level\u0026mdash;from 60\u0026ndash;98.7% for small variants and from 1.9\u0026ndash;84.6% for SVs\u0026mdash;while only slightly increasing the switch error rate for small variants (by 2.7-fold, see Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eA) and the flip error rate for SVs (by 2.3-fold, see Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eA). Compared to statistical phasing without long-read-based scaffold, phasing error rates were reduced significantly for both small variants and SVs, with 12.7-fold reduction of SERs for small variants and 3.6-fold in flip rate for SVs (see Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eA). Compared to the published SV-included haplotype panels (see Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e), both small variants and SVs were phased with error rates lower than statistical-phasing of unrelated individuals alone \u003csup\u003e8,11\u003c/sup\u003e. As such, this hybrid phasing strategy provides an efficient and feasible solution for future reference panel construction.\u003c/p\u003e \u003cp\u003eAs the commonly used resource in imputation and GWAS, the value of this newly constructed haplotype reference panel (Han-SV panel) in imputation was demonstrated with a similar number of high-quality imputed SNVs, compared to the commonly used multi-ancestry expanded-1KGP panel and Chinese population reference panel ChinaMAP. In contrast, Han-SV panel imputed 2.3 or 4.3-fold more SVs with high-quality, compared to the expanded-1KGP panel, on a generated test variant set and SNP array data from Li et al., 2022 \u003csup\u003e27\u003c/sup\u003e. Evaluating the imputation accuracy of high-quality imputed variants, our Han-SV panel achieved similar or higher F1 scores for all variant types across different MAF bins, compared to ChinaMAP and the expanded-1KGP panel. In addition, imputation using this SV-enriched reference panel provides a more practical approach to include SVs in GWAS. In two GWASs of fingerprints and skin-related phenotypes implemented in this study, we not only identified SVs in previously known associated regions but also uncovered unreported associated regions when SVs were included in the GWAS analysis. For example, we identified a fingerprint trait associated DEL located in the prompter of \u003cem\u003eTCF4\u003c/em\u003e in a known associated region (see Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eA), while the lead SNP in this region is identified from SNP-only GWAS located in the intronic region (see Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eB). We also identified a previously unknown independent INS associated skin moisture phenotype, located in the exonic region of two isoforms of \u003cem\u003eLY6S\u003c/em\u003e (see Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eA). Our findings include both independent associated SVs and SVs in linkage disequilibrium with previously reported associated variants, highlighting the value of SV-included GWAS to offer new insights into the genetic basis of complex traits in humans and the strong need of including SVs into future GWAS.\u003c/p\u003e \u003cp\u003eIn conclusion, we presented a Han Chinese haplotype reference panel, including both small variants and SVs, and its high imputation performance allows GWAS to be implemented without additional sequencing costs. This haplotype panel therefore provided a practical and cost-effective solution for the robust analysis of common SVs in the Chinese population, which will strongly advance our knowledge of the genetic underpinnings of disease and complex phenotypes.\u003c/p\u003e"},{"header":"Materials and Methods","content":"\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\n \u003ch2\u003eLRS and SRS data generation\u003c/h2\u003e\n \u003cp\u003eThis study included 943 Han Chinese individuals, recruited from Zhengzhou City in central China. DNA was extracted from whole blood samples, as described previously \u003csup\u003e19\u003c/sup\u003e. In brief, short-read WGS was performed using the BGI DNBSEQ-T7 platform and long-read WGS was performed with Oxford Nanopore Technologies on the PromethION platform. Short-read sequencing reads were aligned to the GRCh38 reference genome using BWA-MEM mode (v0.7.17) with default parameters \u003csup\u003e49\u003c/sup\u003e, resulting in \u0026gt;\u0026thinsp;30X depth of coverage per sample. Oxford Nanopore sequencing reads were aligned to GRCh38 reference gnome using NGMLR (v0.2.7) with default parameters \u003csup\u003e50\u003c/sup\u003e. The LRS mean depth of coverage of 100 samples was 28X (range: 22X \u0026minus;\u0026thinsp;34X) and 843 samples was 14X (range: 10X \u0026minus;\u0026thinsp;26X).\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\n \u003ch2\u003eGenome-wide variant calling and genotyping\u003c/h2\u003e\n \u003cp\u003eSmall variants (including SNV and InDel) were identified based on SRS data using Genome Analysis Toolkit (v4.1.8.1) following the best practice for germline variant calling \u003csup\u003e51,52\u003c/sup\u003e.\u003c/p\u003e\n \u003cp\u003eStructural variants were called individually based on ONT data using sniffles (v2.2) \u003csup\u003e53\u003c/sup\u003e, requiring output names of all supporting reads for each SV (--output-rnames). The tandem repeat annotations for the human reference genome (sniffles accompanied annotation file: human_GRCh38_no_alt_analysis_set.trf.bed) was provided as an additional input to improve SV calling in the repetitive regions in GRCh38. We only included autosomal SVs with PASS in the FILTER field in VCF in the further analyses. We further excluded TRAs, SVs with size less than 50 bp, SVs in telomere, centromere or gap regions or very large SVs with length greater than one third of the chromosome \u003csup\u003e12\u003c/sup\u003e. We implemented SV genotyping individually using LRcaller (v1.0) with default parameters \u003csup\u003e17\u003c/sup\u003e. To obtain higher accuracy in genotyping, we filled the sequences in GRCh38 affected by SV in the REF and ALT fields before LRcaller genotyping to increase genotype accuracy.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\n \u003ch2\u003eBenchmarking samples\u003c/h2\u003e\n \u003cp\u003eTo evaluate the accuracy of phasing methods, we included samples of HG002 from the Genome in a Bottle (GIAB) Consortium \u003csup\u003e23,24\u003c/sup\u003e, and LCL6 from the Chinese Quartet Genome Project \u003csup\u003e25\u003c/sup\u003e, both of which have established benchmark small variant and SV call sets. The benchmark small variant set (NISTv4.2.1) of HG002 was downloaded from GIAB (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://ftp-trace.ncbi.nlm.nih.gov/ReferenceSamples/giab/release/AshkenazimTrio/HG002_NA24385_\u003c/span\u003e\u003c/span\u003e\u003cbr\u003e\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003eson/NISTv4.2.1/GRCh38/HG002_GRCh38_1_22_v4.2.1_benchmark.vcf.gz\u003c/span\u003e\u003c/span\u003e). The benchmark small variant and SV sets of LCL6 and her parents (LCL7 and LCL8) were downloaded from the Chinese Quartet Genome Project (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://chinese-quartet.org\u003c/span\u003e\u003c/span\u003e\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e).\u003c/span\u003e\u003c/p\u003e\n \u003cp\u003eThe benchmark haplotype-resolved small variant and SV set (CMRG_v1.00) of HG002 based on GRCh38 were downloaded from GIAB (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://ftp-trace.ncbi.nlm.nih.gov/ReferenceSamples/giab/release/AshkenazimTrio/HG002_NA24385_son/CMRG_v1.00/GRCh38/Additional\u003c/span\u003e\u003c/span\u003e file 2:Files/HG002v11-align2-GRCh38/HG002v11-align2-GRCh38.dip.vcf.gz), which was used as truth set for phasing accuracy evaluation. To generate a haplotype-resolved variant set of LCL6, we implemented WhatsHap (v1.4) phasing in pedigree mode \u003csup\u003e54\u003c/sup\u003e based on downloaded benchmark small variant and SV call sets of the trio, respectively.\u003c/p\u003e\n \u003cp\u003eTo include HG002 and LCL6 in our phasing pipeline, their ONT sequencing data were obtained as follows. The ONT sequencing data in FASTQ format of HG002 were downloaded from GIAB (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://ftp-trace.ncbi.nlm.nih.gov/ReferenceSamples/giab/data/AshkenazimTrio/HG002_NA24385_\u003c/span\u003e\u003c/span\u003e\u003cbr\u003e\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003eson/UCSC_Ultralong_OxfordNanopore_Promethion/GM24385_1.fastq.gz\u003c/span\u003e\u003c/span\u003e), and mapped to GRCh38 using NGMLR (v0.2.7) with default parameters \u003csup\u003e50\u003c/sup\u003e. The average depth of coverage of HG002 was 40X. The ONT sequencing alignment data (aligned by NGMLR) of LCL6 and her parents (LCL7 and LCL8) were downloaded from the Chinese Quartet Genome Project (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003eftp://download.big.ac.cn/gsa-human/HRA001859/HRR796651\u003c/span\u003e\u003c/span\u003e). Due to the coverage of these samples near 100X, we downsampled them to 27X, 38X, and 39X respectively. The SV calling and genotyping of HG002 and LCL6 were implemented the same as described above based on ONT data.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\n \u003ch2\u003eIntegrating long-read-based and statistical-based phasing for small variants and SVs\u003c/h2\u003e\n \u003cp\u003eThe benchmark small variant datasets from HG002 and LCL6 were merged with the small variant set of our Han Chinese cohort using the bcftools \u003cem\u003emerge\u003c/em\u003e module (v1.20) \u003csup\u003e55\u003c/sup\u003e to create a comprehensive multi-sample variant set. Before phasing, the following filters were applied: genotype missing rate\u0026thinsp;\u0026lt;\u0026thinsp;5%, Hardy-Weinberg Equilibrium (HWE) exact test P-value\u0026thinsp;\u0026gt;\u0026thinsp;1 \u0026times; 10⁻\u003csup\u003e10\u003c/sup\u003e, and allele count\u0026thinsp;\u0026gt;\u0026thinsp;0. The resulting multi-sample VCF was then split into individual VCFs to enable read-based phasing at the individual level.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\n \u003ch2\u003eIndividual-level long-read-based phasing\u003c/h2\u003e\n \u003cp\u003eSmall variants and SVs for each sample in our Han Chinese cohort, as well as for HG002 and LCL6, were co-phased using LongPhase (v1.5.1) with the sequencing platform specified (--ont) and parameters provided for co-phasing InDels and SVs (--indels and --sv-file) \u003csup\u003e21\u003c/sup\u003e. To compare the performance of the read-based co-phasing method with SNV-only phasing, long-read-based phasing of small variants for HG002 and LCL6 was also performed using WhatsHap phase (v1.4) with default settings \u003csup\u003e54\u003c/sup\u003e.\u003c/p\u003e\n \u003cp\u003eAll phased small variant VCF per individual from longphase were merged as one multi-sample VCF file using bcftools \u003cem\u003emerge\u003c/em\u003e (v1.20) \u003csup\u003e55\u003c/sup\u003e. We used Truvari \u003cem\u003ecollapse\u003c/em\u003e (v4.1.0) \u003csup\u003e56\u003c/sup\u003e, a tool designed for SV merging, to combine the phased SVs from the individual SV VCFs, requiring minimum percentage of allele sequence and size similarity as 0.5, and other comparison threshold arguments as default.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec16\" class=\"Section2\"\u003e\n \u003ch2\u003ePopulation-level statistical phasing\u003c/h2\u003e\n \u003cp\u003eSeveral quality control measures for SVs were applied before population-level statistical phasing. This includes removing variants lacking exact sequences in the REF or ALT fields of the VCF, such as INSs without resolved inserted sequences. To prevent a single variant from being classified as both an InDel and an SV in the same sample, InDels larger than 50 bp were excluded. The multi-sample SV set was re-genotyped using LRCaller (v1.0) with default parameters \u003csup\u003e17\u003c/sup\u003e, before statistical phasing to reduce the genotype missing rate caused by SV merging at individual level. Multi-sample set of phased small variants and SVs, as generated through long-read-based phasing (as described above), were then merged using bcftools \u003cem\u003econcat\u003c/em\u003e (v1.20) \u003csup\u003e55\u003c/sup\u003e to construct a haplotype scaffold. Statistical phasing was performed using this scaffold with SHAPEIT5 \u003cem\u003ephase_common\u003c/em\u003e (v5.1.1), employing the genetic map files for GRCh38 provided by SHAPEIT5 \u003csup\u003e22\u003c/sup\u003e. The resulting haplotype set was subsequently filtered using the following criteria: genotype missing rate\u0026thinsp;\u0026lt;\u0026thinsp;5%, HWE exact test P-value\u0026thinsp;\u0026gt;\u0026thinsp;1 \u0026times; 10⁻\u003csup\u003e10\u003c/sup\u003e, and allele count\u0026thinsp;\u0026gt;\u0026thinsp;0.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec17\" class=\"Section2\"\u003e\n \u003ch2\u003eHaplotype phasing performance evaluation\u003c/h2\u003e\n \u003cp\u003eThe phasing accuracy of the small variant and structural variant (SV) haplotype sets for HG002 and LCL6 was evaluated by comparing them to their respective benchmark haplotype sets.\u003c/p\u003e\n \u003cp\u003eFor small variants, phasing accuracy was quantified using the switch error rate (SER), calculated with the WhatsHap \u003cem\u003ecompare\u003c/em\u003e module (v1.4) \u003csup\u003e54\u003c/sup\u003e.\u003c/p\u003e\n \u003cp\u003eFor SVs, phasing accuracy was measured by the flip error rate, which represents the proportion of SVs with phase flips based on SV-flanking-SNV pairs, as previously described by Byrska-Bishop et al., 2022 and Kosugi et al., 2023 \u003csup\u003e8,11\u003c/sup\u003e. Briefly, we calculated the switch error for each phased SV and its nearest small variants on both sides, comparing our phased set to the benchmark haplotype set. Concordant SVs between our phased and benchmark sets were defined as having matching SV types and breakpoints within 10 bp. Concordant small variants identification was computed using bcftools \u003cem\u003eisec\u003c/em\u003e (v1.20) to identify small variants with exact same position in phased and benchmark sets \u003csup\u003e55\u003c/sup\u003e. A flip error was defined as the presence of two switch errors \u003csup\u003e8,11\u003c/sup\u003e.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec18\" class=\"Section2\"\u003e\n \u003ch2\u003eImputation performance evaluation\u003c/h2\u003e\n \u003cp\u003eWe generated a test dataset by randomly selecting 100,000 SNVs from chromosome 20 across 100 randomly selected samples from the Han Chinese cohort. Imputation was performed using Beagle (v5.4) with default parameters \u003csup\u003e57\u003c/sup\u003e, utilizing the haplotype set from the remaining 843 samples as the test Han-SV panel (see \u003cstrong\u003eSupplementary Fig.\u0026nbsp;3\u003c/strong\u003e). To compare the imputation performance to other imputation panels, the expanded-1KGP panel and ChinaMAP were also used as reference to impute the test dataset. The expanded-1KGP panel was obtained from \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://ftp.1000genomes.ebi.ac.uk/vol1/ftp/data_collections/1000G_2504_high_coverage/working/20220422_3202_phased_SNV_indel_SV\u003c/span\u003e\u003c/span\u003e, and imputation was performed using Beagle (v5.4) with default parameters \u003csup\u003e57\u003c/sup\u003e. The ChinaMAP imputation was performed using its imputation server (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://www.mbiobank.com/imputation/\u003c/span\u003e\u003c/span\u003e\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e)\u003c/span\u003e with default settings \u003csup\u003e26\u003c/sup\u003e.\u003c/p\u003e\n \u003cp\u003eTo estimate imputation accuracy, the imputed variants on chromosome 20 of the 100 samples were compared to the variant sets derived from WGS data as truth sets (see \u003cstrong\u003eSupplementary Fig.\u0026nbsp;3\u003c/strong\u003e). For small variants, we compared the imputed genotypes of each sample with the ones derived from the SRS data of the same sample. The imputed true positive small variants were defined as having the same genomic position and reference and alternative alleles as that in SRS data. For SVs, we compared the imputed SVs of each sample with the ones derived from LRS data of the sample sample. The individual LRS-derived SV set of the 100 samples were merged using \u003cem\u003etruvari collapse\u003c/em\u003e (v4.1.0) \u003csup\u003e56\u003c/sup\u003e, requiring minimum percentage of allele sequence and size similarity as 0.5 and other comparison threshold parameters as default. The 100-sample SV set was re-genotyped using LRCaller (v1.0) with default parameters \u003csup\u003e17\u003c/sup\u003e to reduce the genotype missing rate caused by SV merging at individual level. The imputed true positive SVs were defined as having the same SV type and breakpoints within 200 bp as that in the LRS-derived SV set. When assessing imputation accuracy for different MAF bins, both imputed variant and truth set were required in the same MAF bin. Recall was calculated as the fraction of true positives in the truth set and precision was calculated as the fraction of true positives in the imputed set. F1 score was calculated as the harmonic mean of recall and precision.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec19\" class=\"Section2\"\u003e\n \u003ch2\u003eImputation and GWAS using SNP array data\u003c/h2\u003e\n \u003cp\u003eSNP array data and ten fingerprint phenotypes of 9,909 individuals, from three Han Chinese cohorts: the Taizhou Longitudinal Study (TZL, n\u0026thinsp;=\u0026thinsp;2,961), the National Survey of Physical Traits (NSPT, n\u0026thinsp;=\u0026thinsp;2,679) and the Jidong cohort Study (JD, n\u0026thinsp;=\u0026thinsp;4,269), were obtained from Li et al., 2022 \u003csup\u003e27\u003c/sup\u003e. The SNP set was filtered using the following criteria: genotype missing rate\u0026thinsp;\u0026lt;\u0026thinsp;5%, HWE exact test P-value\u0026thinsp;\u0026gt;\u0026thinsp;1 \u0026times; 10⁻\u003csup\u003e10\u003c/sup\u003e and allele count\u0026thinsp;\u0026gt;\u0026thinsp;0. Since the SNP array data was based on GRCh37, we converted the genomic coordinates to GRCh38 using LiftoverVcf from Picard tool (v2.27.5) (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://broadinstitute.github.io/picard\u003c/span\u003e\u003c/span\u003e).\u003c/p\u003e\n \u003cp\u003eImputation was performed using our Han-SV haplotype reference panel with Beagle (v5.4) with default parameters \u003csup\u003e57\u003c/sup\u003e. To compare the imputation performance of Han-SV panel and the expanded-1KGP panel on SNP array data, the expanded-1KGP panel was also used as reference to impute TZL SNP array data on chromosome 20 using Beagle (v5.4) with default parameters \u003csup\u003e57\u003c/sup\u003e.\u003c/p\u003e\n \u003cp\u003eQuality control was applied before GWAS, including removing variants with low imputation quality (dosage r\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.7 from Beagle results) and rare variants with minor allele frequency\u0026thinsp;\u0026lt;\u0026thinsp;0.01 and requiring genotyping missing rate\u0026thinsp;\u0026gt;\u0026thinsp;0.2. The genome-wide association analyses on ordinal phenotypes (coded as 0, 1, and 2 for arch, loop, and whorl, respectively) of ten fingers were performed using PLINK (v2.0) \u003csup\u003e58\u003c/sup\u003e, using multiple linear regression model of additive allelic effects with additional 4 genetic principal components (PCs), 5 genetic PCs, and 5 genetic PCs as covariates in TZL, NSPT, and JD cohorts, respectively as previously described in the original study \u003csup\u003e27\u003c/sup\u003e. We used METAL (version 2011-03-25) \u003csup\u003e59\u003c/sup\u003e to perform fingerprint meta-analyses across the three independent cohorts and selected genome-wide significant variants (P-meta\u0026thinsp;\u0026lt;\u0026thinsp;1.67 \u0026times; 10⁻\u003csup\u003e8\u003c/sup\u003e) after multiple-testing adjustment \u003csup\u003e27\u003c/sup\u003e for further analysis, and we further defined the marginally significance as 1.67 \u0026times; 10⁻\u003csup\u003e8\u003c/sup\u003e \u0026lt; P-meta\u0026thinsp;\u0026lt;\u0026thinsp;3.33 \u0026times; 10⁻\u003csup\u003e6\u003c/sup\u003e.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec20\" class=\"Section2\"\u003e\n \u003ch2\u003eImputation and GWAS using SRS-derived variants\u003c/h2\u003e\n \u003cp\u003eAn independent cohort including 1,016 Han Chinese individuals were recruited from Shanghai city in China. As previously described \u003csup\u003e60\u003c/sup\u003e, short-read WGS was performed using the BGI-DNBSEQ-T7 platform. For each sample, we obtained at least 90 Gb data. Short-read sequencing reads were aligned to the GRCh38 reference genome using BWA-MEM mode (v0.7.17) with default parameters \u003csup\u003e49\u003c/sup\u003e. Small variants (SNV and InDel) were identified using Genome Analysis Toolkit (v4.1.7.0) following the best practice for germline variant calling \u003csup\u003e51,52\u003c/sup\u003e. Only biallelic SNVs with PASS in FILTER field of VCF were included for imputation. Imputation was performed using the Han-SV panel with Beagle (v5.4) with default parameters \u003csup\u003e57\u003c/sup\u003e.\u003c/p\u003e\n \u003cp\u003eThe imputed variants were firstly filtered based on imputation quality (dosage r\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;\u0026ge;\u0026thinsp;0.7), provided by Beagle software \u003csup\u003e57\u003c/sup\u003e. The imputed and called variants were also filtered using PLINK (v1.90 beta 6) \u003csup\u003e58\u003c/sup\u003e specifying minimum MAF (\u0026ndash;maf 0.01) and maximal missing rate (\u0026ndash;geno 0.2). Genome-wide association analysis was performed for 145 skin-related traits (\u003cstrong\u003esee Supplementary Data 8\u003c/strong\u003e) in 1,016 samples on the filtered SV and SNP/InDel markers using the Linear Mixed-Model Association program (EMMAX) \u003csup\u003e61\u003c/sup\u003e with sex, age, and first 6 PCs as covariates. We defined genome-wide significance as P-value\u0026thinsp;\u0026lt;\u0026thinsp;5 \u0026times; 10⁻\u003csup\u003e8\u003c/sup\u003e and the suggestive significance as 5 \u0026times; 10⁻\u003csup\u003e8\u003c/sup\u003e \u0026lt; P-value\u0026thinsp;\u0026lt;\u0026thinsp;1 \u0026times; 10⁻\u003csup\u003e5\u003c/sup\u003e.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec21\" class=\"Section2\"\u003e\n \u003ch2\u003eVariant annotation of significant loci in GWAS\u003c/h2\u003e\n \u003cp\u003eTo identify independent variants, a stepwise selection was conducted using GCTA-COJO (v1.94.1) with default parameters \u003csup\u003e28\u003c/sup\u003e. We defined association regions as genomic regions within 250kb up- and down-stream of the independent variants, and merged the overlapped genomic regions to obtain association regions \u003csup\u003e27\u003c/sup\u003e. Variants within association regions and with LD r\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;\u0026ge;\u0026thinsp;0.3 of the independent variants were defined as candidate variants \u003csup\u003e62,63\u003c/sup\u003e.\u003c/p\u003e\n \u003cp\u003eAll candidate SVs were annotated using AnnotSV (v3.4) with default settings \u003csup\u003e33\u003c/sup\u003e. All candidate SNPs/InDels were annotated by SnpEff (v5.2a) \u003csup\u003e64\u003c/sup\u003e, using the GRCh38.105 database as canonical transcripts (-canon). To annotate variants in non-coding regions of the genome, regulatory region annotations were obtained from the GeneHancer database \u003csup\u003e40\u003c/sup\u003e. SNPs and InDels were identified overlapping regulatory elements if presenting within a\u0026thinsp;\u0026plusmn;\u0026thinsp;200 bp flanking region, using \u003cem\u003ebedtools intersect\u003c/em\u003e \u003csup\u003e65\u003c/sup\u003e. SVs were considered to overlap regulatory regions if either breakpoint was within 200 bp of the regulatory elements.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec22\" class=\"Section2\"\u003e\n \u003ch2\u003eA comparison to the association regions in the EBI GWAS category\u003c/h2\u003e\n \u003cp\u003eWe downloaded the association file from the NHGRI-EBI GWAS catalogue (release version r2023-06-17) \u003csup\u003e32\u003c/sup\u003e and extracted results for skin redness-related phenotypes (hemoglobin measurement), skin disease-related phenotypes (psoriasis, melanoma, keratinocyte cancers), pigmentation-related phenotypes (hair, skin, and eye color), and radiation response (see \u003cstrong\u003eSupplementary Data 9\u003c/strong\u003e). Association regions of skin phenotypes identified in this study were defined as previously reported regions if they overlapped with the variants reported in the NHGRI-EBI GWAS catalogue \u003csup\u003e32\u003c/sup\u003e.\u003c/p\u003e\n\u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe study complied with all relevant regulations for working with human subjects in China. The Ethics Committee of the School of Life Sciences, Fudan University, Shanghai, China approved the study. Participants were recruited to a project studying physical anthropology diversity in China funded by the Ministry of Science and Technology of the People\u0026rsquo;s Republic of China (2015FY111700). Informed consents were approved by all participants.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of data and materials\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe use of Han-SV haplotype reference panel is available through the developed imputation server (https://www.biosino.org/svrp), that can be used to impute both small and structural variants.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eWe thank Dr. Marta Byrska-Bishop for providing the detailed guide to calculate SV phasing flip error rate, and the help from Professor Leming Shi and Dr. Luyao Ren for accessing the Zhonghua Quartet Project data. We also like to acknowledge the support from the Human Phenome Data Center at Fudan University.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis work was supported by grants from the National Key R\u0026amp;D Program of China (2024YFE011610), Noncommunicable Chronic Diseases-National Science and Technology Major Project (2024ZD0533101, and 2024ZD0533100), National Natural Science Foundation of China (grant No. 32370686), and Shanghai Municipal Science and Technology (grant No. 2023SHZDZX02A13) to S.F. and the 111 Project (B13016) to S.F and L.J.. T.G was supported by the Office of China Postdoctoral Council International Postdoctoral Exchange Program Fellowship (YJ20210053).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor contributions:\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eS.F. and L.J. conceived the study. T.G. performed reference panel construction and the evaluation. Y.Z. performed the GWAS applications. Y.H., J.Z. and J.G. provided additional supports in Oxford Nanopore sequencing data processing and J.L. provided supports in GWAS analysis. J.L., Q.P. and S.W. contributed to generating the SNP array and fingerprint phenotype data. T.G. and Y.Z. did the data analysis and interpretation. H.C., L.Y. and G.Z. developed the imputation server. T.G., Y.Z. and S.F., wrote the manuscript. S.F. and L.J. supervised the study. All authors reviewed the manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConflict of interest statement\u003c/strong\u003e:\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eNone declared.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eSudmant, P. H. \u003cem\u003eet al.\u003c/em\u003e An integrated map of structural variation in 2,504 human genomes. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e526\u003c/strong\u003e, 75\u0026ndash;81 (2015).\u003c/li\u003e\n\u003cli\u003eChiang, C. \u003cem\u003eet al.\u003c/em\u003e The impact of structural variation on human gene expression. \u003cem\u003eNat. Genet.\u003c/em\u003e \u003cstrong\u003e49\u003c/strong\u003e, 692\u0026ndash;699 (2017).\u003c/li\u003e\n\u003cli\u003eChaisson, M. J. P. \u003cem\u003eet al.\u003c/em\u003e Multi-platform discovery of haplotype-resolved structural variation in human genomes. \u003cem\u003eNat. Commun.\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, 1784 (2019).\u003c/li\u003e\n\u003cli\u003eMukamel, R. E. \u003cem\u003eet al.\u003c/em\u003e Protein-coding repeat polymorphisms strongly shape diverse human phenotypes. \u003cem\u003eScience\u003c/em\u003e \u003cstrong\u003e373\u003c/strong\u003e, 1499\u0026ndash;1505 (2021).\u003c/li\u003e\n\u003cli\u003eWheeler, M. M. \u003cem\u003eet al.\u003c/em\u003e Whole genome sequencing identifies structural variants contributing to hematologic traits in the NHLBI TOPMed program. \u003cem\u003eNat Commun\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 7592 (2022).\u003c/li\u003e\n\u003cli\u003eBillingsley, K. J. \u003cem\u003eet al.\u003c/em\u003e Genome-Wide Analysis of Structural Variants in Parkinson Disease. \u003cem\u003eAnn Neurol\u003c/em\u003e \u003cstrong\u003e93\u003c/strong\u003e, 1012\u0026ndash;1022 (2023).\u003c/li\u003e\n\u003cli\u003eKaivola, K. \u003cem\u003eet al.\u003c/em\u003e Genome-wide structural variant analysis identifies risk loci for non-Alzheimer\u0026rsquo;s dementias. \u003cem\u003eCell Genom\u003c/em\u003e \u003cstrong\u003e3\u003c/strong\u003e, 100316 (2023).\u003c/li\u003e\n\u003cli\u003eByrska-Bishop, M. \u003cem\u003eet al.\u003c/em\u003e High-coverage whole-genome sequencing of the expanded 1000 Genomes Project cohort including 602 trios. \u003cem\u003eCell\u003c/em\u003e \u003cstrong\u003e185\u003c/strong\u003e, 3426\u0026ndash;3440.e19 (2022).\u003c/li\u003e\n\u003cli\u003eHehir-Kwa, J. Y. \u003cem\u003eet al.\u003c/em\u003e A high-quality human reference panel reveals the complexity and distribution of genomic structural variants. \u003cem\u003eNat. Commun.\u003c/em\u003e \u003cstrong\u003e7\u003c/strong\u003e, 12989 (2016).\u003c/li\u003e\n\u003cli\u003eValls-Margarit, J. \u003cem\u003eet al.\u003c/em\u003e GCAT|Panel, a comprehensive structural variant haplotype map of the Iberian population from high-coverage whole-genome sequencing. \u003cem\u003eNucleic Acids Res.\u003c/em\u003e \u003cstrong\u003e50\u003c/strong\u003e, 2464\u0026ndash;2479 (2022).\u003c/li\u003e\n\u003cli\u003eKosugi, S. \u003cem\u003eet al.\u003c/em\u003e Detection of trait-associated structural variations using short-read sequencing. \u003cem\u003eCell Genom.\u003c/em\u003e \u003cstrong\u003e3\u003c/strong\u003e, 100328 (2023).\u003c/li\u003e\n\u003cli\u003eAudano, P. A. \u003cem\u003eet al.\u003c/em\u003e Characterizing the major structural variant alleles of the human genome. \u003cem\u003eCell\u003c/em\u003e \u003cstrong\u003e176\u003c/strong\u003e, 663\u0026ndash;675.e19 (2019).\u003c/li\u003e\n\u003cli\u003eEbert, P. \u003cem\u003eet al.\u003c/em\u003e Haplotype-resolved diverse human genomes and integrated analysis of structural variation. \u003cem\u003eScience\u003c/em\u003e \u003cstrong\u003e372\u003c/strong\u003e, eabf7117 (2021).\u003c/li\u003e\n\u003cli\u003eGlusman, G., Cox, H. C. \u0026amp; Roach, J. C. Whole-genome haplotyping approaches and genomic medicine. \u003cem\u003eGenome Med.\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, 73 (2014).\u003c/li\u003e\n\u003cli\u003eZhang, X., Wu, R., Wang, Y., Yu, J. \u0026amp; Tang, H. Unzipping haplotypes in diploid and polyploid genomes. \u003cem\u003eComput. Struct. Biotechnol. J.\u003c/em\u003e \u003cstrong\u003e18\u003c/strong\u003e, 66\u0026ndash;72 (2020).\u003c/li\u003e\n\u003cli\u003eGarg, S. Computational methods for chromosome-scale haplotype reconstruction. \u003cem\u003eGenome Biol.\u003c/em\u003e \u003cstrong\u003e22\u003c/strong\u003e, 101 (2021).\u003c/li\u003e\n\u003cli\u003eBeyter, D. \u003cem\u003eet al.\u003c/em\u003e Long-read sequencing of 3,622 Icelanders provides insight into the role of structural variants in human diseases and other traits. \u003cem\u003eNat. Genet.\u003c/em\u003e \u003cstrong\u003e53\u003c/strong\u003e, 779\u0026ndash;786 (2021).\u003c/li\u003e\n\u003cli\u003eWu, Z. \u003cem\u003eet al.\u003c/em\u003e Structural variants in the Chinese population and their impact on phenotypes, diseases and population adaptation. \u003cem\u003eNat. Commun.\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 6501 (2021).\u003c/li\u003e\n\u003cli\u003eGong, J. \u003cem\u003eet al.\u003c/em\u003e Long-read sequencing of 945 Han individuals identifies structural variants associated with phenotypic diversity and disease susceptibility. \u003cem\u003eNat Commun\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, 1494 (2025).\u003c/li\u003e\n\u003cli\u003eCretu Stancu, M. \u003cem\u003eet al.\u003c/em\u003e Mapping and phasing of structural variation in patient genomes using nanopore sequencing. \u003cem\u003eNat. Commun.\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, 1326 (2017).\u003c/li\u003e\n\u003cli\u003eLin, J.-H., Chen, L.-C., Yu, S.-C. \u0026amp; Huang, Y.-T. LongPhase: an ultra-fast chromosome-scale phasing algorithm for small and large variants. \u003cem\u003eBioinformatics\u003c/em\u003e \u003cstrong\u003e38\u003c/strong\u003e, 1816\u0026ndash;1822 (2022).\u003c/li\u003e\n\u003cli\u003eHofmeister, R. J., Ribeiro, D. M., Rubinacci, S. \u0026amp; Delaneau, O. Accurate rare variant phasing of whole-genome and whole-exome sequencing data in the UK Biobank. \u003cem\u003eNat. Genet.\u003c/em\u003e \u003cstrong\u003e55\u003c/strong\u003e, 1243\u0026ndash;1249 (2023).\u003c/li\u003e\n\u003cli\u003eWagner, J. \u003cem\u003eet al.\u003c/em\u003e Benchmarking challenging small variants with linked and long reads. \u003cem\u003eCell Genom.\u003c/em\u003e \u003cstrong\u003e2\u003c/strong\u003e, 100128 (2022).\u003c/li\u003e\n\u003cli\u003eWagner, J. \u003cem\u003eet al.\u003c/em\u003e Curated variation benchmarks for challenging medically relevant autosomal genes. \u003cem\u003eNat. Biotechnol.\u003c/em\u003e \u003cstrong\u003e40\u003c/strong\u003e, 672\u0026ndash;680 (2022).\u003c/li\u003e\n\u003cli\u003eRen, L. \u003cem\u003eet al.\u003c/em\u003e Quartet DNA reference materials and datasets for comprehensively evaluating germline variant calling performance. \u003cem\u003eGenome Biol.\u003c/em\u003e \u003cstrong\u003e24\u003c/strong\u003e, 270 (2023).\u003c/li\u003e\n\u003cli\u003eLi, L. \u003cem\u003eet al.\u003c/em\u003e The ChinaMAP reference panel for the accurate genotype imputation in Chinese populations. \u003cem\u003eCell Res\u003c/em\u003e \u003cstrong\u003e31\u003c/strong\u003e, 1308\u0026ndash;1310 (2021).\u003c/li\u003e\n\u003cli\u003eLi, J. \u003cem\u003eet al.\u003c/em\u003e Limb development genes underlie variation in human fingerprint patterns. \u003cem\u003eCell\u003c/em\u003e \u003cstrong\u003e185\u003c/strong\u003e, 95\u0026ndash;112.e18 (2022).\u003c/li\u003e\n\u003cli\u003eYang, J. \u003cem\u003eet al.\u003c/em\u003e Conditional and joint multiple-SNP analysis of GWAS summary statistics identifies additional variants influencing complex traits. \u003cem\u003eNat. Genet.\u003c/em\u003e \u003cstrong\u003e44\u003c/strong\u003e, 369\u0026ndash;75, S1\u0026ndash;3 (2012).\u003c/li\u003e\n\u003cli\u003eSepp, M., Pruunsild, P. \u0026amp; Timmusk, T. Pitt-Hopkins syndrome-associated mutations in TCF4 lead to variable impairment of the transcription factor function ranging from hypomorphic to dominant-negative effects. \u003cem\u003eHum. Mol. Genet.\u003c/em\u003e \u003cstrong\u003e21\u003c/strong\u003e, 2873\u0026ndash;2888 (2012).\u003c/li\u003e\n\u003cli\u003eLehalle, D., Williams, C., Siu, V. M. \u0026amp; Clayton-Smith, J. Fetal pads as a clue to the diagnosis of Pitt-Hopkins syndrome. \u003cem\u003eAm. J. Med. Genet. A\u003c/em\u003e \u003cstrong\u003e155A\u003c/strong\u003e, 1685\u0026ndash;1689 (2011).\u003c/li\u003e\n\u003cli\u003ePitt, D. \u0026amp; Hopkins, I. A syndrome of mental retardation, wide mouth and intermittent overbreathing. \u003cem\u003eAust. Paediatr. J.\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 182\u0026ndash;184 (1978).\u003c/li\u003e\n\u003cli\u003eSollis, E. \u003cem\u003eet al.\u003c/em\u003e The NHGRI-EBI GWAS Catalog: knowledgebase and deposition resource. \u003cem\u003eNucleic Acids Res.\u003c/em\u003e \u003cstrong\u003e51\u003c/strong\u003e, D977\u0026ndash;D985 (2023).\u003c/li\u003e\n\u003cli\u003eGeoffroy, V. \u003cem\u003eet al.\u003c/em\u003e AnnotSV: an integrated tool for structural variations annotation. \u003cem\u003eBioinformatics\u003c/em\u003e \u003cstrong\u003e34\u003c/strong\u003e, 3572\u0026ndash;3574 (2018).\u003c/li\u003e\n\u003cli\u003eGray, M., Bliss, D. Z. \u0026amp; McNichol, L. Moisture-Associated Skin Damage: Expanding and Updating Practice Based on the Newest ICD-10-CM Codes. \u003cem\u003eJ. Wound Ostomy Continence Nurs.\u003c/em\u003e \u003cstrong\u003e49\u003c/strong\u003e, 143\u0026ndash;151 (2022).\u003c/li\u003e\n\u003cli\u003eShmerling, M. \u003cem\u003eet al.\u003c/em\u003e a New IFN-Inducible Human Member of the Ly6a Subfamily Expressed by Spleen Cells and Associated with Inflammation and Viral Resistance. \u003cem\u003eImmunohorizons\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, 253\u0026ndash;272 (2022).\u003c/li\u003e\n\u003cli\u003eTu, Z. \u003cem\u003eet al.\u003c/em\u003e Pro-inflammatory cytokine IL-6 regulates LMO4 expression in psoriatic keratinocytes via AKT/STAT3 pathway. \u003cem\u003eImmun Inflamm Dis\u003c/em\u003e \u003cstrong\u003e11\u003c/strong\u003e, e1104 (2023).\u003c/li\u003e\n\u003cli\u003eBoehncke, W.-H. \u0026amp; Sch\u0026ouml;n, M. P. Psoriasis. \u003cem\u003eLancet\u003c/em\u003e \u003cstrong\u003e386\u003c/strong\u003e, 983\u0026ndash;994 (2015).\u003c/li\u003e\n\u003cli\u003eHalprin, K. M. Epidermal \u0026lsquo;turnover time\u0026rsquo;--a re-examination. \u003cem\u003eBr. J. Dermatol.\u003c/em\u003e \u003cstrong\u003e86\u003c/strong\u003e, 14\u0026ndash;19 (1972).\u003c/li\u003e\n\u003cli\u003eNestle, F. O., Kaplan, D. H. \u0026amp; Barker, J. Psoriasis. \u003cem\u003eN. Engl. J. Med.\u003c/em\u003e \u003cstrong\u003e361\u003c/strong\u003e, 496\u0026ndash;509 (2009).\u003c/li\u003e\n\u003cli\u003eFishilevich, S. \u003cem\u003eet al.\u003c/em\u003e GeneHancer: genome-wide integration of enhancers and target genes in GeneCards. \u003cem\u003eDatabase \u003c/em\u003e \u003cstrong\u003e2017\u003c/strong\u003e, (2017).\u003c/li\u003e\n\u003cli\u003eNishikori, S. \u003cem\u003eet al.\u003c/em\u003e Resistance training rejuvenates aging skin by reducing circulating inflammatory factors and enhancing dermal extracellular matrices. \u003cem\u003eSci. Rep.\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 10214 (2023).\u003c/li\u003e\n\u003cli\u003eIchihashi, M. \u0026amp; Ando, H. The maximal cumulative solar UVB dose allowed to maintain healthy and young skin and prevent premature photoaging. \u003cem\u003eExp. Dermatol.\u003c/em\u003e \u003cstrong\u003e23 Suppl 1\u003c/strong\u003e, 43\u0026ndash;46 (2014).\u003c/li\u003e\n\u003cli\u003eLy, B. C. K., Dyer, E. B., Feig, J. L., Chien, A. L. \u0026amp; Del Bino, S. Research Techniques Made Simple: Cutaneous Colorimetry: A Reliable Technique for Objective Skin Color Measurement. \u003cem\u003eJ. Invest. Dermatol.\u003c/em\u003e \u003cstrong\u003e140\u003c/strong\u003e, 3\u0026ndash;12.e1 (2020).\u003c/li\u003e\n\u003cli\u003evan Beek, N. \u003cem\u003eet al.\u003c/em\u003e Thyroid hormones directly alter human hair follicle functions: anagen prolongation and stimulation of both hair matrix keratinocyte proliferation and hair pigmentation. \u003cem\u003eJ. Clin. Endocrinol. Metab.\u003c/em\u003e \u003cstrong\u003e93\u003c/strong\u003e, 4381\u0026ndash;4388 (2008).\u003c/li\u003e\n\u003cli\u003eCzarnywojtek, A. \u003cem\u003eet al.\u003c/em\u003e The role of antithyroglobulin, antiperoxidase and anti-TSH receptor autoantibodies in amiodarone-induced thyrotoxicosis and amiodarone-induced hypothyroidism (A two-center study). \u003cem\u003eNeuro Endocrinol. Lett.\u003c/em\u003e \u003cstrong\u003e36\u003c/strong\u003e, 677\u0026ndash;681 (2015).\u003c/li\u003e\n\u003cli\u003eYan, X. \u003cem\u003eet al.\u003c/em\u003e Peroxidasin is essential for eye development in the mouse. \u003cem\u003eHum. Mol. Genet.\u003c/em\u003e \u003cstrong\u003e23\u003c/strong\u003e, 5597\u0026ndash;5614 (2014).\u003c/li\u003e\n\u003cli\u003eChoi, Y., Chan, A. P., Kirkness, E., Telenti, A. \u0026amp; Schork, N. J. Comparison of phasing strategies for whole human genomes. \u003cem\u003ePLoS Genet.\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, e1007308 (2018).\u003c/li\u003e\n\u003cli\u003eBrowning, S. R. \u0026amp; Browning, B. L. Haplotype phasing: existing methods and new developments. \u003cem\u003eNat Rev Genet\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 703\u0026ndash;714 (2011).\u003c/li\u003e\n\u003cli\u003eLi, H. Aligning sequence reads, clone sequences and assembly contigs with BWA-MEM. \u003cem\u003earXiv [q-bio.GN]\u003c/em\u003e (2013) doi:10.48550/ARXIV.1303.3997.\u003c/li\u003e\n\u003cli\u003eSedlazeck, F. J. \u003cem\u003eet al.\u003c/em\u003e Accurate detection of complex structural variations using single-molecule sequencing. \u003cem\u003eNat. Methods\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, 461\u0026ndash;468 (2018).\u003c/li\u003e\n\u003cli\u003ePoplin, R. \u003cem\u003eet al.\u003c/em\u003e Scaling accurate genetic variant discovery to tens of thousands of samples. \u003cem\u003ebioRxiv\u003c/em\u003e (2017) doi:10.1101/201178.\u003c/li\u003e\n\u003cli\u003eVan der Auwera, G. A. \u0026amp; O\u0026rsquo;Connor, B. D. \u003cem\u003eGenomics in the Cloud: Using Docker, GATK, and WDL in Terra\u003c/em\u003e. (O\u0026rsquo;Reilly Media, 2020).\u003c/li\u003e\n\u003cli\u003eSmolka, M. \u003cem\u003eet al.\u003c/em\u003e Detection of mosaic and population-level structural variants with Sniffles2. \u003cem\u003eNat. Biotechnol.\u003c/em\u003e \u003cstrong\u003e42\u003c/strong\u003e, 1571\u0026ndash;1580 (2024).\u003c/li\u003e\n\u003cli\u003eMartin, M. \u003cem\u003eet al.\u003c/em\u003e WhatsHap: fast and accurate read-based phasing. \u003cem\u003ebioRxiv\u003c/em\u003e (2016) doi:10.1101/085050.\u003c/li\u003e\n\u003cli\u003eDanecek, P. \u003cem\u003eet al.\u003c/em\u003e Twelve years of SAMtools and BCFtools. \u003cem\u003eGigascience\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, (2021).\u003c/li\u003e\n\u003cli\u003eEnglish, A. C., Menon, V. K., Gibbs, R. A., Metcalf, G. A. \u0026amp; Sedlazeck, F. J. Truvari: refined structural variant comparison preserves allelic diversity. \u003cem\u003eGenome Biol.\u003c/em\u003e \u003cstrong\u003e23\u003c/strong\u003e, 271 (2022).\u003c/li\u003e\n\u003cli\u003eBrowning, B. L., Tian, X., Zhou, Y. \u0026amp; Browning, S. R. Fast two-stage phasing of large-scale sequence data. \u003cem\u003eAm. J. Hum. Genet.\u003c/em\u003e \u003cstrong\u003e108\u003c/strong\u003e, 1880\u0026ndash;1890 (2021).\u003c/li\u003e\n\u003cli\u003ePurcell, S. \u003cem\u003eet al.\u003c/em\u003e PLINK: a tool set for whole-genome association and population-based linkage analyses. \u003cem\u003eAm. J. Hum. Genet.\u003c/em\u003e \u003cstrong\u003e81\u003c/strong\u003e, 559\u0026ndash;575 (2007).\u003c/li\u003e\n\u003cli\u003eWiller, C. J., Li, Y. \u0026amp; Abecasis, G. R. METAL: fast and efficient meta-analysis of genomewide association scans. \u003cem\u003eBioinformatics\u003c/em\u003e \u003cstrong\u003e26\u003c/strong\u003e, 2190\u0026ndash;2191 (2010).\u003c/li\u003e\n\u003cli\u003eGong, J. \u003cem\u003eet al.\u003c/em\u003e Long-read sequencing of 945 Han individuals identifies novel structural variants associated with phenotypic diversity and disease susceptibility. \u003cem\u003emedRxiv\u003c/em\u003e (2024) doi:10.1101/2024.03.21.24304654.\u003c/li\u003e\n\u003cli\u003eKang, H. M. \u003cem\u003eet al.\u003c/em\u003e Variance component model to account for sample structure in genome-wide association studies. \u003cem\u003eNat. Genet.\u003c/em\u003e \u003cstrong\u003e42\u003c/strong\u003e, 348\u0026ndash;354 (2010).\u003c/li\u003e\n\u003cli\u003eRamos, E. \u003cem\u003eet al.\u003c/em\u003e Replication of genome-wide association studies (GWAS) loci for fasting plasma glucose in African-Americans. \u003cem\u003eDiabetologia\u003c/em\u003e \u003cstrong\u003e54\u003c/strong\u003e, 783\u0026ndash;788 (2011).\u003c/li\u003e\n\u003cli\u003eLee, J. J. \u003cem\u003eet al.\u003c/em\u003e Gene discovery and polygenic prediction from a genome-wide association study of educational attainment in 1.1 million individuals. \u003cem\u003eNat. Genet.\u003c/em\u003e \u003cstrong\u003e50\u003c/strong\u003e, 1112\u0026ndash;1121 (2018).\u003c/li\u003e\n\u003cli\u003eCingolani, P. \u003cem\u003eet al.\u003c/em\u003e A program for annotating and predicting the effects of single nucleotide polymorphisms, SnpEff: SNPs in the genome of Drosophila melanogaster strain w1118; iso-2; iso-3. \u003cem\u003eFly \u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, 80\u0026ndash;92 (2012).\u003c/li\u003e\n\u003cli\u003eQuinlan, A. R. \u0026amp; Hall, I. M. BEDTools: a flexible suite of utilities for comparing genomic features. \u003cem\u003eBioinformatics\u003c/em\u003e \u003cstrong\u003e26\u003c/strong\u003e, 841\u0026ndash;842 (2010).\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"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":"nature-portfolio","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"","title":"Nature Portfolio","twitterHandle":"","acdcEnabled":false,"dfaEnabled":false,"editorialSystem":"ejp","reportingPortfolio":"","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Reference Panel, Structural Variant, Haplotype Phasing, Long-Read Sequencing, Genome-Wide Association Study","lastPublishedDoi":"10.21203/rs.3.rs-6312842/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6312842/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eHaplotype reference panels are commonly used for genotype imputation in genome-wide association studies (GWAS). Although structural variations (SVs) are recognized as major contributors to human phenotypes, they are often excluded from GWAS analyses. Here, we integrate long-read-based and statistical methods to provide a comprehensive haplotype reference panel (Han-SV panel) incorporating 32,603,300 single nucleotide variants (SNPs), 3,180,227 small deletions and insertions and 172,569 SVs derived from 943 Han Chinese individuals. Our hybrid phasing approach had a 12.7-fold reduction in phasing error for small variants and 3.6-fold for SVs compared to conventional statistical phasing. This Han-SV panel enabled a more than two-fold in amount and four-fold in accuracy improvement of SV imputation compared to the expanded 1000 Genomes Project panel. Two GWASs using our panel-imputed variants identified 69 associated SVs and 101 previously unreported regions associated with skin-related and fingerprint phenotypes\u0026mdash;substantially outperforming both short-read and SNP-array-based GWAS. This Han-SV panel offers a valuable resource for variant imputation and SV-included association studies to further uncover the novel phenotype associations and address critical gaps in missing heritability. An imputation server was provided for the use of the Han-SV panel (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.biosino.org/svrp\u003c/span\u003e\u003cspan address=\"https://www.biosino.org/svrp\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e).\u003c/p\u003e","manuscriptTitle":"A Long-read based Haplotype Panel Enhances Imputation and Discovery of Functional Small and Structural Variants","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-04-10 05:06:32","doi":"10.21203/rs.3.rs-6312842/v1","editorialEvents":[],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"nature-communications","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"NCOMMS","sideBox":"Learn more about [Nature Communications](http://www.nature.com/ncomms/)","snPcode":"","submissionUrl":"https://mts-ncomms.nature.com/","title":"Nature Communications","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Nature Communications","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"1fda6238-e2f0-4414-8002-8efbf22d3710","owner":[],"postedDate":"April 10th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":46875983,"name":"Biological sciences/Computational biology and bioinformatics"},{"id":46875984,"name":"Biological sciences/Genetics/Genomics"}],"tags":[],"updatedAt":"2025-10-04T05:50:08+00:00","versionOfRecord":[],"versionCreatedAt":"2025-04-10 05:06:32","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6312842","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6312842","identity":"rs-6312842","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}