Ethnic minority groups in China

A pangenome reference of 36 Chinese populations

Abstract
Human genomics is witnessing an ongoing paradigm shift from a single reference sequence to a pangenome form, but populations of Asian ancestry are underrepresented. Here we present data from the first phase of the Chinese Pangenome Consortium, including a collection of 116 high-quality and haplotype-phased de novo assemblies based on 58 core samples representing 36 minority Chinese ethnic groups. With an average 30.65× high-fidelity long-read sequence coverage, an average contiguity N50 of more than 35.63 megabases and an average total size of 3.01 gigabases, the CPC core assemblies add 189 million base pairs of euchromatic polymorphic sequences and 1,367 protein-coding gene duplications to GRCh38. We identified 15.9 million small variants and 78,072 structural variants, of which 5.9 million small variants and 34,223 structural variants were not reported in a recently released pangenome reference1. The Chinese Pangenome Consortium data demonstrate a remarkable increase in the discovery of novel and missing sequences when individuals are included from underrepresented minority ethnic groups. The missing reference sequences were enriched with archaic-derived alleles and genes that confer essential functions related to keratinization, response to ultraviolet radiation, DNA repair, immunological responses and lifespan, implying great potential for shedding new light on human evolution and recovering missing heritability in complex disease mapping.

Main
Over the past two decades, the reference human genome sequence has served as the foundation for genetic and biomedical research and applications; however, there is a broad consensus that no single reference sequence can represent the genomic diversity of global populations. On one hand, high-quality population-specific and haplotype-resolved genome references are necessary for genetic and medical analysis2. On the other hand, there is a clear need to shift from a single reference to a pangenome form that better represents genomic diversity, or allelic variation, within and across human populations3. With the advancement of long-read sequencing technologies as well as computational methods, it is now feasible to enable pan-genomic construction to capture the missed variations from a large collection of diverse genomes4. The Human Pangenome Reference Consortium (HPRC) recently constructed a draft human pangenome reference based on 47 samples of worldwide populations, but with East Asian population samples underrepresented (n = 4)1. In particular, only three Southern Han Chinese (CHS) samples were included in the HPRC reference, too few to represent the genomic diversity of ethnic groups in a region such as China, which is populated by 1.44 billion people. We showed previously that the genetic diversity in Asia was not well covered by large-scale international collaborative projects such as the 1000 Genomes Project5,6. Although the need to improve the representation of diverse ancestral backgrounds in genomic research is well known7,8, substantially fewer genomic studies have been conducted in populations of Asian ancestry compared with populations of European ancestry. China harbours a great genetic diversity, with 55 officially recognized minority ethnic groups in addition to the Han Chinese majority and a considerable number of unrecognized ethnic groups. Despite advances in sequencing technologies leading to the achievement of the telomere-to-telomere haploid assembly T2T-CHM13 (ref. 9), only a limited number of Chinese genomes have been de novo assembled to high-quality haplotype sequences using long-read DNA sequencing technologies2,10,11,12,13,14. The only two published studies on the Chinese pangenome were limited to the short-read sequencing data of Han Chinese samples15,16. There is an urgent need to establish a high-quality pangenome reference that better represents the great genomic diversity of Chinese populations. We anticipate such an effort to broaden the reference to represent genomic diversity, resolve allelic and locus heterogeneity, support unbiased and comprehensive detection of structural variation within and across populations, and improve genotyping accuracy in genomic regions enriched with complex sequence variations, such as human leukocyte antigen genes, and ultimately facilitate genomic analysis for both evolutionary and medical research.

The Chinese Pangenome Consortium (CPC) aims to de novo sequence at least 500 individuals to better detect and catalogue sequence variants of the Chinese populations using third-generation sequencing technologies. Here we present the draft Chinese pangenome reference based on the first sequencing effort (Phase I) of the CPC, reporting 116 high-quality de novo assemblies from 58 core samples representing 36 minority Chinese ethnic groups and 6 assemblies of the Han Chinese majority.

Genomic features of the CPC assemblies
To investigate the genomic composition of our 58 CPC samples, we aligned the 116 assemblies to T2T-CHM13 and called the variants of each sample using Phased Assembly Variant caller17 (Methods). We observed an average of 11.22% (9.35–13.05%) small variants and 24.16% (19.72–29.3%) structural variants (SVs) per CPC sample were in the highly repetitive regions (Extended Data Fig. 2), and more than half of the SVs were singletons except for those on the Y chromosome (Supplementary Fig. 6). We further identified an average of 2,802 (2,265–3,187) novel sequences for each CPC sample by collecting insertions larger than 1 kb compared to the T2T-CHM13 reference (Methods and Extended Data Fig. 3a). These novel sequences were located at 63,243 regions across the genome, and they were particularly enriched in the highly repetitive sequences such as the centromeric and telomeric regions (Extended Data Fig. 3b). We determined 115 genomic regions as insertion hotspots based on the insertion frequency, covering a total length of 204.8 Mb.

Considering that reference bias typically hinders the identification of novel SVs in both short- and long-read datasets, we further completed the analysis of complex structural variations (CSVs) using our CPC high-quality and phased haplotype sequences. On average, we identified 70 CSVs per sample with the newly developed method SVision18 (Supplementary Fig. 7). The most common CSV type was INS+tDUP with an average number of 44 (Supplementary Fig. 8). We merged the CSVs into a set of 706 non-redundant CSVs (Extended Data Fig. 4a) and classified them into four classes19: shared (all samples), major (≥50% of samples), polymorphic (<50% of samples) and singleton (only one sample). We found that the peak of the CSV length distribution was at 10 kb (Extended Data Fig. 4b). We found one shared CSV (chr11:66,245,313–66,246,097 with an INS+INV type), 34 major CSVs, 277 polymorphic CSVs and 394 singleton CSVs from the CPC assembly set (Extended Data Fig. 4c).

We next annotated the copy number variations (CNVs) relative to the GRCh38 reference in each assembly (Methods). There were 1,367 protein-coding genes in the full set of assemblies that had a gain in copy number in at least one genome. An average of 53 (27–100) genes with a gain in copy number were observed in each assembly (Fig. 2a), 13.39% of CNV genes showed a frequency of more than 5% in the whole CPC assembly set, and 57.86% of CNV genes were found in only a single haplotype. We found there were 1,079 duplicated genes in the CPC assembly set that were not observed in HPRC assemblies (Fig. 2b and Supplementary Table 4). These genes were enriched significantly in olfactory-related functions (for example, GO:0007608, sensory perception of smell, odds ratio (OR) = 6.46, Benjamini–Hochberg (BH)-adjusted P = 2.22 × 10−36; and hsa04740, olfactory transduction, OR = 7.14, BH-adjusted P = 8.35 × 10−38), and at a marginal significance level in skin-related disease (for example, DOID:37, skin disease, OR = 2.33, BH-adjusted P = 0.086; and DOID:16, integumentary system disease, OR = 2.23, BH-adjusted P = 0.086; Supplementary Table 5). We found that 562 (52.1%) of the novel duplicated genes in the CPC assembly set were trait-associated according to the genome-wide association study (GWAS) catalogue, and 207 (19.2%) of these genes were related to at least one human disease annotated by the Disease Ontology, suggesting the remarkable potential for the CPC assembly set to contribute to disease and phenotype association studies. In particular, KCNJ18 and MTRNR2L6 showed considerably high frequency in the CPC assemblies (115 of 116) but were absent in the HPRC assemblies (Fig. 2c). KCNJ18 encodes a member of the inwardly rectifying potassium channel family (Kir2.6), which is expressed in skeletal muscle and is transcriptionally regulated by thyroid hormone20. Kir2.6 alterations were reported to be associated with thyrotoxic periodic paralysis, which is a well-known complication of thyrotoxicosis in Asian populations, but is rare in European populations21,22,23. The pseudogene MTRNR2L6 is predicted to be involved in the regulation of apoptosis (GO:1900118). Notably, this gene was reported to show signatures of positive selection across geographical Han Chinese populations24. A number of the CPC-specific CNV genes showed signatures of natural selection indicated by the estimates of Tajima’s D (Supplementary Fig. 9 and Supplementary Table 4). Outstanding examples include SPATA31D4 (nhaplotype = 6, Tajima’s D = −3.01, false-discovery rate (FDR)-adjusted P = 2.38 × 10−5) and PSAPL1 (nhaplotype =  7, Tajima’s D = 4.32, FDR-adjusted P = 0.006). In particular, SPATA31D4 belongs to the core duplicon families that are thought to have contributed significantly to hominoid evolution with a strong signal of positive selection in hominoids. Previous studies suggested that the gene family has a function in response to ultraviolet radiation and DNA repair, and may also influence human lifespan25.

Fig. 2: CNVs identified from CPC assemblies.



a, Number of duplicated protein-coding genes per CPC genome assembly compared with the GRCh38 reference. b, Venn diagram showing the number of duplicated genes in CPC, HPRC.EAS and HPRC.nEAS assemblies. c, The top 20 most common CPC-specific CNV-related genes compared with the HPRC assemblies. d, The five overlapped CNV genes showing a higher frequency (≥5%) in CPC assemblies (blue) than in HPRC assemblies (orange). HPRC.EAS, East Asian in HPRC; HPRC.nEAS, non-East Asian in HPRC.

There were 288 duplicated genes from the CPC assemblies that were also observed in the HPRC assemblies, of which 123 genes were shared with 4 HPRC East Asian samples (HPRC.EAS), and 278 genes were shared with the remaining 40 non-East Asian HPRC (HPRC.nEAS) samples (Fig. 2b). Among these overlapped genes, we found several genes with a higher frequency (≥5%) in the CPC assembly set than in the HPRC assemblies (Fig. 2d and Supplementary Table 6). For example, the CNV frequency at CYP2D6 was higher in the CPC assembly set than those in the collective HPRC assemblies and in the separate sets of HPRC.EAS or HPRC.nEAS (Fig. 2d and Extended Data Fig. 5). CYP2D6 is well known for its remarkable polymorphism, which has been systematically surveyed in a wide variety of human populations26. We also found CFC1 with highly differentiated CNVs between the CPC and HPRC assemblies (frequency difference = 0.124). Several investigations reported associations between CFC1 and cardiovascular disease27, especially in Chinese populations28. In addition, CTAG1A and ZNF658 were identified with rare CNVs (around 1%) in the HPRC assembly but presented with common CNVs (≥5% and <95%) in the CPC assembly. CTAG1A is expressed at high levels in the normal ovary and testis. ZNF658 regulates the transcription of genes involved in zinc homoeostasis and affects ribosome biogenesis29, and is involved in infectious processes (hsa05168, herpes simplex virus 1 infection).