Research report: The seeds of tomorrow: SEAD reference panel promotes genomics in Asian human populations

A new commentary published in Frontiers in Genetics examines how a large Asian-focused genetic reference panel could reshape the fairness and power of genome-wide association studies, or GWAS, for roughly 60 percent of the world’s population. The article, written by Lidiya Ashenafi and David Karasik of Bar-Ilan University and Hebrew SeniorLife, is a Perspective piece rather than a new data analysis, but it draws together the evidence behind the South and East Asian reference database, known as SEAD, and argues that ancestry-matched reference haplotypes are becoming essential tools for discovering genetic variants that global panels routinely miss.

The logic behind imputation is central to the story. A GWAS depends on the density of variants available for analysis, and many of the variants that matter are rare and specific to particular ancestries. Imputation fills in untyped genetic variants from array or low-coverage sequencing data, increasing the number of variants that can be analyzed and boosting statistical power. The problem, as the authors describe it, is that the largest public imputation reference panels, such as the 1000 Genomes Project and TOPMed, are skewed toward European ancestries. That imbalance means imputation accuracy is uneven across human populations, a gap the authors frame as a fairness issue for genomic medicine.

SEAD was built to address that gap. According to the abstract, the panel derives from more than 11,000 individuals drawn from across Asian countries, providing over 22,000 haplotypes, two per individual, with representation spanning 17 ethnicities. The panel was constructed by integrating existing multi-population resources, including the Westlake BioBank for Chinese, SG10K, GAsP, and the 1000 Genomes Project. By combining these sources, SEAD increased both reference haplotype size and geographic span relative to the individual component panels, and improved the imputation of rare and low-frequency variants in several evaluated Asian populations, with especially strong coverage of individuals from China, Singapore, and India.

That strength comes with an important caveat that the commentary emphasizes. Despite the breadth of the panel, the sample distribution is heavily skewed: China and Singapore contribute more than 80 percent of all individuals, while many other East and South Asian populations contribute far fewer haplotypes. Substantial gaps remain, with populations such as Taiwanese, Myanmar, Bhutan, the Maldives, and Nepal absent or poorly represented in existing reference resources. The authors caution that geographic coverage should not be interpreted as equivalent representation across Asia, particularly because rare variant imputation depends strongly on the number of ancestry-matched haplotypes available for the target population.

The performance evidence is nuanced. SEAD demonstrated improved imputation particularly for rare and low-frequency variants relative to the evaluated 1000 Genomes, TOPMed, and ChinaMAP panels in selected East and South Asian datasets, and for most South and Central Asian populations evaluated it outperformed all three comparators. In East Asian populations, SEAD outperformed TOPMed and 1000 Genomes but did not consistently beat ChinaMAP, which showed higher accuracy in Han Chinese samples, while SEAD performed similarly in Japanese samples and better in Cambodian samples. The authors also note that the comparators comprised two global panels and one Chinese-specific panel, and that SEAD was not benchmarked against a large dedicated South Asian resource such as GAsP2. The takeaway is that panel benefit depends on both size and genetic similarity to the target population, and that one broad Asian panel is not universally better than intensely sampled population-specific panels.

Common variants are a different matter. The evidence for improved common variant imputation with SEAD is less conclusive than for rare and low-frequency variants, because the performance analyses were primarily restricted to variants with minor allele frequency below 5 percent, with the highest reported frequency bin at 2 to 5 percent. Common variants with minor allele frequency of at least 5 percent were therefore not evaluated separately. Given the measured genetic distance between East and South Asian populations, with an FST of 0.06 reported from an analysis of 1000 Genomes samples, the authors argue that population-specific rather than broadly continental reference panels, combined with fair representation from both groups, are likely to improve rare variant imputation in particular.

The commentary also reviews what SEAD-enabled analysis has actually found. Researchers performed GWAS on two correlated bone mineral density traits, at the hip and femoral neck, using data from the Westlake BioBank for Chinese collection imputed with SEAD. The study identified both common and rare variant associations with these traits. The common variant signals occurred at FMN2 and near SOX4, both previously reported for heel, total-body, or femoral-neck bone mineral density, providing confirmation of established BMD loci. Beyond these, the study reported a previously unreported rare variant association near the Syntrophin gamma 1 gene, SNTG1, for hip bone mineral density. Importantly, the same lead variant was also recovered after imputation with ChinaMAP, WBBC-seq, and SG10K, with comparable allele frequencies, effect estimates, and genotype counts, so the signal was not unique to SEAD imputation.

The significance of the SNTG1 finding, the authors argue, lies in improved detectability of a rare variant signal in an ancestry-matched cohort rather than in demonstrating that SNTG1 variation is exclusive to East Asian populations. The variants at this locus are rare in Non-Finnish European and South Asian populations but more common in African American populations, illustrating population-specific allele frequency patterns. Differences in allele frequency and ancestry-specific linkage disequilibrium patterns affect imputation accuracy, particularly for rare variants, and can influence statistical power and the ability to detect or replicate associations across populations. A previous study in a Hispanic cohort also linked the SNTG1 locus to osteoporosis through a different low-frequency variant, rs72645876, and the Musculoskeletal Knowledge Portal reports that common variants in the locus are associated with heel bone density in Europeans, although that association did not reach genome-wide significance.

The researchers went further than association mapping by performing functional experiments on the newly identified SNTG1 variant, carrying out overexpression experiments in mouse preosteoblast cells, the MC3T3-E1 line. The commentary stresses the limits of this work: the experiments were restricted to in vitro assays in a single cell type, with no in vivo or multi-cellular experiments, so the results indicate a hypothesis about the gene’s role in bone formation rather than established biology. The authors note that replication in other populations, together with broader functional approaches, would strengthen the case for SNTG1’s role in bone metabolism.

Looking outward, the commentary situates SEAD within a growing family of Asian imputation resources. The BioBank Japan 1 KG+7K panel improved low-frequency and rare variant imputation relative to TOPMed and enabled serum uric acid and total cholesterol associations that had not reached genome-wide significance under TOPMed imputation. In South Asia, the GAsP2 panel improved coronary artery disease polygenic risk prediction compared with 1000 Genomes-based imputation. A Southeast Asian-specific panel improved the number and quality of imputed variants in ancestry-matched Orang Asli samples despite its small size, and the VN1K panel, built from 1,011 Kinh Vietnamese individuals, outperformed both 1000 Genomes and SG10K in Vietnamese test samples. These results show that population-specific reference haplotypes can improve recovery of rare and locally enriched variants even when the ancestry-matched panel is smaller than a global one, leading the authors to recommend that population-specific and broad regional panels be treated as complementary and that future studies benchmark ancestry-matched panels against broader ones, especially when rare variants are the focus.

The broader message is one of unfinished work. Central Asia and the Middle East remain poorly represented in existing resources as of the end of 2025, and newer resources, including the Northeast Asian Reference Database, an Indian reference panel, the Singapore Precision Health long-read sequencing project, VN1K, and the First Arab Human Pangenome, could help close the gaps. For a continent containing approximately 60 percent of the world’s population, the authors conclude, further studies are essential to enhance genomic coverage, and unequal representation within reference panels may continue to influence imputation accuracy, particularly for rare and population-enriched variants, until that representation becomes genuinely equitable.

Subject of Research: An Asian-focused genetic reference panel for improving genome-wide association study imputation of rare variants.

Article Title: The seeds of tomorrow: SEAD reference panel promotes genomics in Asian human populations

Article References: Ashenafi, L., & Karasik, D. (2026). The seeds of tomorrow: SEAD reference panel promotes genomics in Asian human populations. Frontiers in Genetics, 17, 1926569. https://doi.org/10.3389/fgene.2026.1926569

Image Credits: AI Generated

DOI: 10.3389/fgene.2026.1926569

Keywords: SEAD, genome-wide association study, genetic imputation, rare variants, Asian populations, reference panel, bone mineral density, SNTG1, population genetics, GWAS, Frontiers in Genetics, ancestry representation