Diversity and Scale: Genetic Architecture of 2,068 Traits in the VA Million Veteran Program

Abstract

Genome-wide association studies (GWAS) have underrepresented individuals from non-European populations, impeding progress in characterizing the genetic architecture and consequences of health and disease traits. To address this, we present a population-stratified phenome-wide GWAS followed by a multi-population meta-analysis for 2,068 traits derived from electronic health records of 635,969 participants in the Million Veteran Program (MVP), a longitudinal cohort study of diverse U.S. Veterans genetically similar to the respective African (121,177), Admixed American (59,048), East Asian (6,702), and European (449,042) superpopulations defined by the 1000 Genomes Project. We identified 38,270 independent variants associating with one or more traits at experiment-wide $\left(P<4.6\times {10}^{-11}\right)$ significance; fine-mapping 6,318 signals identified from 613 traits to single-variant resolution. Among these, a third (2,069) of the associations were found only among participants genetically similar to non-European reference populations, demonstrating the importance of expanding diversity in genetic studies. Our work provides a comprehensive atlas of phenome-wide genetic associations for future studies dissecting the architecture of complex traits in diverse populations.

Fig. 1. Overview of the study population, genetic association results, and post-GWAS findings.

The top panel describes the demographic characteristics of the study population. The middle panel features two scatter plots: the left plot illustrates the number of participants (y-axis) with each binary trait grouped by category(x-axis), and the right plot presents the sample (y-axis) across quantitative traits (x-axis). Distinct colors and shapes represent population groups. The bottom panel is organized into three sections: the left section summarizes the study data, the middle section provides key metrics of GWAS results, such as the count of independent loci and lead SNPs, and the right section briefly outlines the post-GWAS findings.

Fig. 2. Multi-population genetic associations with traits.

Combined multi-trait Manhattan plots and bar plots summarizing 98,485 associations with lead SNPs for binary and quantitative traits (p-value<4.6 × 10⁻¹¹). Manhattan plots for (a) binary traits and (b) quantitative traits display associations across chromosomes (x-axis) and −log10 P values (y-axis). Circles represent previously reported associations, while triangles indicate novel trait associations. The size of the triangles corresponds to the effect size, with upward triangles denoting risk associations and downward triangles signifying protective associations. On the top, gene names indicate previously reported variant-trait associations (in black) and new trait associations (in pink). Stacked bar plot for (c) binary traits and (d) quantitative traits enumerate the number of associations with lead SNPs across different trait categories. The left panel presents the count of known associations (green), novel variants-trait associations (blue), and novel variants (light blue). Trait categories are ordered by the number of lead SNPs in descending order. The right panel is a dodged bar plot highlighting associations with lead SNPs based on their MAF categories: common variants (lower opacity) and low-frequency variants (higher opacity). The distribution of known associations (green) and novel SNPs (light blue) is shown for each trait category.

Fig. 3. Multi-Population Fine-Mapped Signals.

(a) Upset plot of cross-population signal sharing for the 57,601 fine-mapped signals. Red colored-portions of bars represent signals where one or more variants show a suggestive association $\left(P<1\times {10}^{-3}\right)$ in an unmapped population, and blue colored-portions represent signals where the unmapped populations were underpowered to detect suggestive associations for any of the variants in the merged approximate credible set. Signal counts are displayed above the bars for intersections in which fewer than 1,000 signals were identified. (b) Scatter plot of the number of signals detected per phenotype vs. the meta-analyzed sample size for the phenotype. Effective sample sizes were used for binary phenotypes, and points are colored by the phenotype category. (c) The distribution of merged approximate credible set sizes for the fine-mapped signals. (d) Coding enrichment across increasingly precise fine-mapped signals. Bars are colored by the proportion of each represented by each grouped Variant Effect Predictor (VEP) annotation, and the black boxes illustrate the proportion of each bar attributable to coding variation. P-values reflect the results of pairwise Wilcoxon tests for coding annotation enrichment. (e) Distribution of effect sizes vs. minor allele frequencies for high-confidence (PIP > 0.95) associations fine-mapped in quantitative (top) and binary (bottom) phenotypes. Each point represents a unique high-confidence variant-phenotype-population mapping. Point colors reflect the population in which they are mapped, and their shapes reflect whether they are a phenotype association previously reported in the GWAS catalog (square), a novel phenotype association for a variant previously reported in the catalog (triangle), or a novel SNP and association not previously reported (circle). Inset bar plots reflect the proportions of high-confidence associations in these three categories across the four tested populations.

Fig. 4. Fine-mapped cross-trait associations.

Chromosome ideogram illustrating highconfidence cross-trait associations (PIP > 0.95) between genetic variants and independent traits. Independence was determined using a phenotypic correlation coefficient threshold of 0.2 and a stepwise selection process accounting for the largest PIP value, meta-analysis p-value, population-specific p-values, effect size, and trait specificity (e.g., gender-specific traits). The ideogram highlights genomic regions associated with three or more traits, including both previously reported associations and novel trait associations discovered in the present study. Colors represent distinct trait categories, while shapes distinguish between previously reported (circle) and novel trait associations (triangle). Each data point on the ideogram indicates that at least one trait within the respective category is associated with the genomic region.