Pathogenic variation distribution of ACMG 3.2 list genes in 11 ethnic groups in southwest China

Abstract

In recent years, human whole genome sequencing projects in China and abroad have made accidental discoveries in different populations. However, the genetic variants present in the multiethnic population of the southwest region of China remain unclear. The purpose of this study is to elucidate the frequency distribution of different types of genetic variations, particularly pathogenic variants, in ACMG v3.2 list genes among 11 ethnic minority populations from the Guizhou Multi-ethnic Genome Database (GMGD). We collected whole-genome sequencing data from 476 individuals across these 11 GMGD populations and classified the effects of the identified variants according to the ACMG/AMP guideline by GeneBe tools [1]. A total of 521 variants were identified after variant filter, of which 1.5%(8/521) of genetic variants were classified as high confidence pathogenic or potentially pathogenic. The most common P/LP variants were ATP7B: rs762866453, ATP7B: rs191312027 and we observed significant population differences. The three novel variants (ATP7B: rs778732681, PALB2 :rs876660147, GAA: rs765718882) exhibited low to moderate allele frequencies (0.106%–0.42%) in the GMGD population but were undetected in mainstream databases. This study provides information on pathogenic/likely pathogenic variations and their frequencies for 73 disease-related genes across 11 ethnic groups, contributing valuable insights for the advancement of genomic medicine. To the best of our knowledge, this represents the first analysis of clinically incidental genetic variants among these 11 ethnic groups and offers further insights into interethnic genetic differences.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12864-026-12934-1.

Introduction

The mapping of human genetic polymorphism through large-scale genome sequencing is of great significance for understanding complex traits and population genomic differences. Genomic studies of diverse populations contribute to building a more comprehensive and equitable medical genomic database [2]. This helps identify genetic variations linked to diseases, predict an individual’s disease risk more accurately, and develop medications tailored to specific genetic profiles [3].

The majority of large-scale human genomics and bioinformatics databases are Westerner-focused, such as the Genome Aggregation Database (gnomAD), the 1000 Genomes Project (1 KGP), and the UK Biobank [4–6]. From 2020, China conducted a series of WGS projects providing vital data supplements for medical genetics [7, 8]. Peng Zhang et al. on the basis of 2999 Chinese Han majority population samples, obtained the NyuWa Genome Database (NCVD) through whole-genome sequencing and ANNOVAR mutation annotation. They found 2,500,000 new variants, 22,504 protein function loss variants and 1,140 pathogenic variants, which were used to study the genetic polymorphisms of the Han population [7]. The China Metabolic Analytics Project (ChinaMap) constructed whole-genome database on basis of 10,588 samples from eight ethnic groups in China, mainly Han population, and published 73,929,146 new variants. Noteworthy, half of 1,661 disease-causing variants were new loci not found in multiple internationally used databases [8]. However, as a multi-ethnic country, many ethnic genetic maps have not been well understood.

In 2015, the American College of Medical Genetics and Genomics (ACMG) and Association of Molecular Pathology (AMP) issued a set of standards for evaluating the pathogenicity of variant loci, which established a framework for Sequence Variation Interpretation (SVI) [9]. In 2019, the ACMG v2.0 updated the list to 56 genes [10]. Most previous studies focused on the frequency distribution of pathogenic and likely pathogenic variants in the 56 genes published by ACMG v2.0, such as Europe (3.4%) [11], Africa (1.2%) [11], Japan (21%) [12], Korea (7%) [13], Qatar (2.3%) [14]. In 2021, ACMG v3.0 updated the list of 73 genes related to Mendelian genetic diseases [15]. All of us analyzed 73 ACMG genes in 98,590 participants and revealed that European participants carried the highest rate of disease-causing variants (2.26%), which was associated mainly with genes associated with breast/ovarian cancer (BRCA1, BRCA2, PALB2) or hypercholesterolemia (APOB, LDLR) [16]. The Westlake BioBank for the Chinese Pilot Project (WBBC) analyzed the distribution of pathogenic variation or possible pathogenic variation loci in 4480 Chinese individuals and found that the highest proportion of pathogenic mutations were TNNT2(24.14%) and MUTYH(24.07%) [17]. Knowing which genes are expected to more frequently carry pathogenic or likely pathogenic variants across different populations is critical for medical genetic research and clinical practice.

Historically, southwest China was a relatively closed region. Due to differences in ethnic cultural background and living habits, ethnic groups rarely intermarried, forming a genetically isolated group. Some ethnic groups have relatively homogenous genetic backgrounds and may even retain some rare genetic traits and genes associated with disease. It is a rare and valuable genetic resource in human population genetics and medical genetics research. We sequenced the genomes of 476 individuals from 11 ethnic groups in southwest region. In total of 12,538,367 variations are obtained of which 11.9% are novel. Following the ACMG/AMP variant classification guidelines, this study estimated clinically actionable variants in 11 ethnic minorities and demonstrated differences in the frequency of pathogenic variants among different ethnic groups. This is the first time that whole genome sequencing has been used to analyze the disease variation of ethnic groups in Southwest China. Our research provides an important supplement to the landscape of pathogenic and likely pathogenic variants in minority populations.

Methods

GMGD study population

A total of 476 healthy, self-declared, and unrelated individuals from diverse ethnic groups in Guizhou Province, Southwest China, voluntarily participated in the GMGD study. These participants, who had no familial relationships, represented 11 ethnic minorities from three language branches: Hmong–Mien languages (Miao, n = 94; Yao, n = 30), Tai-Kadai languages (Dong, n = 40; Bouyei, n = 53; Gelao, n = 39; Zhuang, n = 30; Sui, n = 40), and Sino-Tibetan languages (Yi, n = 40; Bai, n = 30; Tujia, n = 39; Hui, n = 40). The genomic sequence variants and their allele frequencies from these participants were previously published by our group through the Guizhou Multi-ethnic Genome Database (GMGD, https://db.cngb.org/pop/gmgd) [18]. This study was conducted in accordance with the Measures for the Ethical Review of Life Science and Medical Research Involving Humans and was approved by the Research Ethics Committee of the Affiliated Hospital of Guizhou Medical University (IRB No. 2021201). Prior to participation, all individuals provided informed consent by completing and signing an informed consent form.

Sequencing and variant calling

The procedures for sequencing library construction, variant calling, and data filtering in this study were consistent with those described in our previous work (see reference). In brief, after obtaining informed consent from all participants, peripheral blood samples were collected and genomic DNA was extracted. The whole-genome sequencing library was constructed as follows: (1) Genomic DNA was randomly fragmented using Covaris, and fragments of 200–400 bp were selected using magnetic beads. (2) DNA fragments were subjected to end repair and 3′ adenylation. Illumina-compatible adaptors were then ligated to the 3′ adenylated fragments using T4 DNA ligase. (3) The adaptor-ligated fragments were amplified by PCR. (4) The resulting libraries were quality-assessed and sequenced on the DNBSEQ-T1 platform using paired-end 100 bp reads at an average depth of 5.5×.

Sequencing reads were aligned to GRCh38 reference genomes using BWA with the paired-end alignment option. After initial quality control, duplicate reads were removed using SAMtools, and base quality score recalibration (BQSR) was performed using GATK (v3.8). After duplicate reads removal and BQSR, variants were called using the Haplotyper algorithm in GVCF mode. Joint genotyping was performed using GVCFtyper to generate a cohort-level VCF file. Variant Quality Score Recalibration (VQSR) was conducted using GATK v4.1.2, with truth sensitivity filter levels set to 99.1% for SNPs and 99.8% for indels.

To ensure high-quality variants, we applied several filtering steps. Variants that did not pass VQSR were excluded. We also removed sites where genotype quality (GQ) was < 10 in more than 50% of samples, and variants with excessive heterozygosity (ExcessHet > 54.69) as calculated by GATK. Additionally, variants located too close to indels were filtered out using bcftools filter with the options --SnpGap 3 and --IndelGap 5.

To assess the accuracy of genotype calling in the GMGD dataset, we compared low-coverage (5.5×) WGS data with high-coverage (30×) data from 34 individuals. High-coverage genotypes were treated as the ground truth. Only variants with a call rate > 95% per sample and a minor allele frequency (MAF) > 1% were included in the analysis. Autosomal variants detected by both low- and high-coverage sequencing were used to calculate non-reference sensitivity and non-reference genotype concordance. The results from our previously published GMGD database article indicate that, assuming high-depth WGS data as the gold standard, we achieved genotype concordance and non-reference (NR) genotype concordance rates of 0.951 and 0.9606, respectively, with increasing sequencing depth.

Variant annotation and classification

For our analysis, we selected a list of 81 actionable genes recommended by ACMG SF v3.2 [19] (Supplementary Table S1), including genes related to cancer, cardiomyopathy, metabolism, and other miscellaneous phenotypes. These variants are annotated via the Ensembl V variant Effect Predictor (VEP, release 100.0), which integrates several database and computation tools for functional annotation. The tool includes RefGene, dbSNP and ClinVar and various global population databases. Examples include the Genome Aggregation Database (gnomAD v3.0) and the 1000 Genomes Project (1 KGP). The functional effects were interpreted via the dbNSFP database, which includes SIFT, polyphen-2 CADD mutation pathogenicity prediction tools. ClinVar (hg38_clinvar_20210501) was used to assess the clinical significance of the variants.

The workflow for identifying high-confidence pathogenic and likely pathogenic (P/LP) variants is illustrated in Fig. 1 and involved the following steps: (1) Variants with a minor allele frequency (MAF) ≥ 0.05 were excluded based on population datasets (1000 Genomes, gnomAD) integrated into ANNOVAR. (2) The remaining variants were filtered to include only those with protein-coding effects (e.g., missense, nonsense) or splicing-altering consequences. (3) Variant pathogenicity was classified according to the ACMG/AMP guidelines using Genebe.net. (4) High-confidence P/LP variants were further validated through cross-annotation using ClinVar and Genebe pathogenicity combenie tools.

Fig. 1.

Schematic diagram of the variation annotation workflow

Comparison of allele frequencies of pathogenic variation

To calculate the allele frequencies of high-confidence pathogenic and likely pathogenic (P/LP) variants, they were compared with the existing population databases 1 KGP, gnomAD v3.0, the NyuWa Chinese Population Variant Database (NCVD) and the Westlake BioBank for Chinese (WBBC).

Results

Overview of genetic variation and Population stratification

We sequenced the whole genomes of 476 ethnic minorities. The average sequencing depth was 5.52×. The average quality scores of Q20 and Q30 were 97.37 and 91.59, respectively. The GATK tool was used to genotype SNP and indel by comparing the human reference genome version GRCh38/hg38, and The GMGD dataset encompasses ~ 16.34 million variants identified from the genomes of 476 unrelated, healthy, and self-declared individuals. Notably, 11.9% (1,956,322) of these variants are novel and not cataloged in the dbSNP database.In the variation map, C:G > T:A was the dominant mutation, followed by T: A > C:G (Supplementary Figure S1).

Through the principal component analysis, we initially constructed the genetic distribution map of GMGD population. On the East Asian scale, GMGD_Miao, GMGD_Yao formed two branches with other populations, indicating possible genetic drift within the two populations (Fig. 2a). At the global scale, there are obvious genetic differences between the GMGD population and other ethnic groups, indicating that the GMGD population may be genetically homogenous due to the influence of geographical environment, cultural customs and other backgrounds (Fig. 2b).

Fig. 2.

Results of 11 GMGD populations principal component analysis. a Based on East Asian regional scale principal component analysis; b Based on world-scale principal component analysis

Annotation of protein sequence changes or splicing variations according to ACMG/AMP guidelines

To analyze high-confidence pathogenic (P) and likely pathogenic (LP) variants, we performed VEP-based annotations of the GMGD dataset using a variant filtering pipeline (Figure. 1). This workflow focused on rare and putatively pathogenic variants associated with ACMG 3.2-listed disease genes. First, 41,274 genetic variants linked to these genes were identified. Subsequently, applying a conservative filtering strategy—allele frequency (AF) < 0.05 in major population datasets (e.g., 1000 Genomes, gnomAD) and protein-coding effects (e.g., missense, nonsense mutations or splicing-altering variants)—we narrowed the dataset to 521 rare variants, as shown in Fig. 3A.

Fig. 3.

Distribution of variants type and across different disease categories

The 521 variants included 316 missense variants, 190 synonymous variants, 6 stop gained/lost variants, 5 stop gained variants, 2 frameshift variants, 1 start lost variants and 7 splice region variant. These variants were subjected to in-depth genetic interpretation using the ACMG/AMP guidelines by classified the GENEbe tool, annotating each variation manually. As shown in Fig. 3B, these variants are mainly associated with Cancer, Cardiovascular, Metabolic, and miscellaneous categories. By searching the disease database Clinvar, GeneBe we found 8 high confidence P/LP variants, with detailed annotations summarized in Table 1. We further used Polyphen-2, SIFT CADD, and AlphaMissense computer prediction software to find a total of 8 P/LP variants. These 8 P/LP variants were predicted to be deleterious with high confidence across different computational tools, as shown in Table 2.

Table 1.

Summary of pathogenic variation and possible pathogenic variation according to ACMG/AMP guideline annotations

Gene	rs ID	Variant class	acmg_classification	acmg_criteria	No. of participants
APOB	rs5742904	missense_variant	Pathogenic	PS1_Moderate&PM5&PP5_Very_Strong&BP4	1
APOB	rs144467873	missense_variant	Pathogenic	PM5&PP5_Very_Strong	1
ATP7B	rs778732681	missense_variant	Pathogenic	PM1&PM2&PM5&PP2&PP3_Strong&PP5_Very_Strong	3
ATP7B	rs762866453	missense_variant	Pathogenic	PM1&PM5&PP2&PP3_Strong&PP5_Very_Strong	20
ATP7B	rs191312027	missense_variant	Pathogenic	PM1&PM5&PP2&PP5_Very_Strong	9
PALB2	rs876660147	frameshift_variant	Pathogenic	PVS1&PM2&PP5_Very_Strong	1
GAA	rs765718882	stop_gained	Pathogenic	PM3&PVS1&PM2&PP4	2
LDLR	rs768563000	missense_variant	Pathogenic	PP3&PP4&PM2&PS4&PM3	2

Table 2.

Summary of functional predictions for pathogenic/likely pathogenic (P/LP) variants

Gene	rs ID	SIFT_score	SIFT_pred	Polyphen2_HDIV_score	Polyphen2_HDIV_pred	CADD_phred	AlphaMissense_score	AlphaMissense_pred
APOB	rs5742904	0.003	D	.	.	26.8	0.161	B
APOB	rs144467873	0	D	.	.	25.3	0.221	B
ATP7B	rs778732681	0.001	D	1	D	14.73	0.915	P
ATP7B	rs762866453	0	D	1	D	24.4	0.661	P
ATP7B	rs191312027	0.001	D	1	D	25.3	0.409	A
PALB2	rs876660147	.	.	.	.	.	.	.
GAA	rs765718882	.	.	.	.	35	.	.
LDLR	rs768563000	0.004	D	0.975	D	25.2	0.477	A

Comparison of allele frequencies of P/LP variants in different population genome datasets

The frequencies of 8 P/LP variant alleles were obtained from 476 minority population datasets and compared with public databases, the results are shown in Table 3. Three of the 8 P/LP variants were new variants, and no population frequencies were found in the 1 KGP and gnomAD databases. The allele frequency of two ATP7B variants showed marked ethnic-specific differences: rs762866453 (p.Gly1287Ser) exhibited an AF of 2.33% in the GMGD population, rs191312027 (p.Gly869Arg) had an AF of 0.955% in GMGD, higher than WBBC (0.134%) and gnomAD_EAS (0.022%) which was significantly higher than those reported in other global databases (1KGP: undetectable, gnomAD_EAS: 0.017%, WBBC: 0.045%). These variants were not detected in 1KGP or gnomAD databases but exhibited low to moderate AF in the GMGD population: rs778732681 (p.Arg1320Ser): 0.42%, rs876660147 (p.Gln921ThrfsTer7): 0.106%, rs765718882 (p.Glu888Ter): 0.211%, These findings suggest these variants may be ethnic-specific to the GMGD minority populations.

Table 3.

Allele frequencies of pathogenic/likely pathogenic (P/LP) variants in GMGD and multiple public databases

Gene	rs ID	GMGD AF	1KGP_AF	gnomAD_EAS_AF	WBBC_AF	NVCD_AF
APOB	rs5742904	0.00105932	.	.	.	0.00016672
APOB	rs144467873	0.00105932	0.001	0.001305	0.001116071	0.0016672
ATP7B	rs778732681	0.00421941	.	.	0.000334821	.
ATP7B	rs762866453	0.0233051	.	0.0001669	0.000446429	0.00050017
ATP7B	rs191312027	0.00955414	0.002	0.0002225	0.001339286	.
PALB2	rs876660147	0.00105932	.	.	0.000223214	.
GAA	rs765718882	0.0021097	.	0.0002719	0.000111607	0.00050017
LDLR	rs768563000	0.00211864	.	0.00005437	.

Discussion

Understanding human genomic variation could help yield insights into the population affinities, disease etiology and genetic basis of phenotypic diversity. Whole genome sequencing analysis of different human populations could help discover novel human variants, calculate genotype frequency baseline, and dissect population genetic structure and development [20–23]. Southwest China is a region with diverse ethnic minority species and stable genetic characteristics, which provides excellent advantages for studying the genetic backgrounds of different ethnic populations. In our study, we collected whole genome data from 11 ethnic minorities and found 11.9% of detected variants were novel. In line with the historical and linguistic grouping, the Miao and Yao ethnic groups are distinguished from other Sino Tibetan language families. unexpectedly, the Miao in our queue did not cluster together as expected with the Miao in the 1KGP. This is similar to previous research findings that the Southwest Hmong–Mien-speaking ethnic group has population separation from the Miao, Yao ethnic group in public database and obvious genetic substructures existed in geographically different Hmong–Mien-speaking populations [24]. Furthermore, we observed significant differences in the prevalence of pathogenic variants among different ethnic groups. Conducting genomic research on 11 ethnic groups is crucial for building a broader and more equitable genomic database.

In 2013, the American College of Medical Genetics and Genomics (ACMG) convened a working group consisting of representatives from ACMG, the Society for Molecular Pathology (AMP), and the College of American Pathologists to develop a set of standards and guidelines for the interpretation of sequence variations, which have been widely followed. With the increase in clinical exome and genome sequencing (ES/GS) data, the ACMG was updated to the ACMG SF3.0 list, which contained 73 genes currently. The use of SF genes in asymptomatic individuals may provide an opportunity to identify potential genetic risk factors [25]. In this study, we analyzed the distribution of P/LP mutation loci in 81 ACMG SF3.2 Mendelian genetic disease-related genes in 476 ethnic minority populations.

Variants in the ATP7B gene are closely associated with Wilson disease (WD), and the observed frequency discrepancies of these variants in this study likely reflect the genetic background of the GMGD population [26]. For instance, the variant rs762866453 (p.Gly1287Ser) exhibited an allele frequency (AF) of 2.33% in the GMGD population, which is significantly higher than that reported in global databases (undetectable in 1KGP, AF = 0.017% in gnomAD_EAS). This marked enrichment may stem from genetic isolation, founder effects, or adaptive selection within the GMGD group. Similarly, rs191312027 (p.Gly869Arg) had an AF of 0.955% in GMGD, over 43 times higher than that in the East Asian population (AF = 0.022% in gnomAD_EAS) and WBBC (AF = 0.045%). These findings underscore the potential deviation of pathogenic variant frequencies in minority populations from global averages, highlighting the risks of relying solely on existing databases for genetic counseling or diagnostic strategies, which could lead to underestimation or overestimation of disease risks.

Three novel variants (rs778732681, rs876660147, rs765718882) were identified in the GMGD population, with low to moderate allele frequencies (0.106–0.42%), yet they were undetected in mainstream databases (1KGP and gnomAD). This discrepancy likely arises from two factors: (1) insufficient representation of minority populations in existing databases, leading to undercapture of low-frequency variants; and (2) the unique genetic architecture of the GMGD group, which harbors “ethnic-specific” variants. These findings emphasize the necessity of independent genetic studies in non-majority populations to enhance the comprehensiveness of global genomic data and mitigate clinical misinterpretations caused by data bias.

Data differences in clinical genomic interpretation are a recognized problem, and clinical testing is guided in part by ancestry. As more evidence (functional analysis, family isolation studies, large case-control studies) builds up, it will be possible to better understand the genetic risk burden in population groups and precisely explain disease-related variants. Our study highlights the need for more population-specific case-control studies in Guizhou to determine the distribution of the described variants across cases and controls, as well as large lineages with complete clinical and genetic examinations to help assess the degree of pathogenicity of each variant. Due to the lack of clinical data, we are unable to track and follow up on individuals carrying pathogenic variants. Further work that is integrated with clinical practice is needed.

Conclusions

Overall, we established a genomic database of 476 individuals from GMGD through whole genome sequencing and obtained a large number of new variants(11.9%). We annotated a total of 8 P/LP variants in the ACMG SF3.2 gene list. The most common P/LP variants were ATP7B: rs762866453, ATP7B: rs191312027 and we observed significant population differences. Notably, as ATP7B is associated with autosomal recessive disease, it is clinically recommended that variants in this gene be reported as secondary findings only when two P/LP variants (biallelic) are detected—a key consideration for interpreting our observed ATP7B variants in the context of clinical application. The three novel variants (ATP7B: rs778732681, PALB2 :rs876660147, GAA: rs765718882) exhibited low to moderate allele frequencies (0.106%–0.42%) in the Guizhou Minority Genetic Diversity (GMGD) population but were undetected in mainstream databases. Our research results included a wider range of populations with different genetic backgrounds, which is crucial for conducting disease variant carrier screening in natural populations for clinical intervention.

Authors' contributions

Y.L. and C.L. both contributed to the study conception and design. Material preparation, data collection, and analysis were performed by S.Z., A.C., F.L., C.W., T.Z., H.Y., S.K., and X.Q. The first draft of the manuscript was written by Y.L., C.L., S.Z., F.L., X.J., and Y.H. both reviewed, edited, and approved the final manuscript.

Funding

Funding was provided by the Natural Science Foundation of China (Grant Nos. 32360154, 32560152, 31560306, 82360813), the Science and Technology Fund of Guizhou Provincial Health Commission (gzwkj2024-056), Guizhou Provincial Science and Technology Program Project (QKH Platform Talents-GCC[2023]035, QKH Platform Talents-CXTD[2023]003, QKH-MS-2026-536), the project of Key Laboratory of Endemic and Ethnic Diseases, Ministry of Education, Guizhou Medical University (No.FZSW-2025-003), Guangzhou Basic and Applied Basic Research Foundation (No.202201010189), the project of Key Laboratory of Endemic and Ethnic Diseases, Ministry of Education, Guizhou Medical University (No.FZSW-2022-001), National Natural Science Foundation of China (No.U25D9022), Key-Area Research and Development Program of Guangdong Province (No.2023B0303040001), Shenzhen Science and Technology Program (No.SYSPG20241211173852024), the Innovation Platform for Academicians of Hainan Province (No.YSPTZX202118).

Data availability

The data that support the findings of this study have been deposited into CNGB Sequence Archive (CNSA) [27] of China National GeneBank DataBase (CNGBdb) [28] with accession number CNP0003074. The variation data reported in this paper have been deposited in the Genome Variation Map (GVM) [29] in National Genomics Data Center, Beijing Institute of Genomics, Chinese Academy of Sciences and China National Center for Bioinformation [30], under accession number GVM000478: https://ngdc.cncb.ac.cn/gvm.

Declarations

Ethics approval and consent to participate

All procedures followed were in accordance with the ethical standards of the responsible committee on human experimentation (institutional and national) and with the Helsinki Declaration of 1975, as revised in 2008 (5). This study was reviewed and approved by the Ethics Committee of Affiliated Hospital of Guiyang Medical College [Ethics 2015 No. (29), Ethics 2023 No. (92)].

All participants signed informed consent.

Competing interests

The authors declare no competing interests.