Cross-population GWAS and proteomics improve risk prediction and reveal mechanisms in atrial fibrillation

Abstract

Atrial fibrillation (AF) is a common cardiac arrhythmia with strong genetic components, yet its underlying molecular mechanisms and potential therapeutic targets remain incompletely understood. We conducted a cross-population genome-wide meta-analysis of 168,007 AF cases and identified 525 loci that met genome-wide significance. Two loci of PITX2 and ZFHX3 genes were identified as shared across populations of different ancestries. Comprehensive gene prioritization approaches reinforced the role of muscle development and heart contraction while also uncovering additional pathways, including cellular response to transforming growth factor-beta. Population-specific genetic correlations uncovered common and unique circulatory comorbidities between Europeans and Africans. Mendelian randomization identified modifiable risk factors and circulating proteins, informing disease prevention and drug development. Integrating genomic data from this cross-population genome-wide meta-analysis with proteomic profiling significantly enhanced AF risk prediction. This study advances our understanding of the genetic etiology of AF while also enhancing risk prediction, prevention strategies, and therapeutic development.

Fig. 1. Study design overview.

AF atrial fibrillation, AFR African, AMR Admixed American, GWAS genome-wide association study. EAS Eastern Asian, EUR European, SAS South Asian.

Fig. 2. Genetic loci associated with atrial fibrillation (AF) across populations of different ancestries.

a Manhattan plot of GWAS associations. The x-axis represents the genomic positions of SNPs across chromosomes, while the y-axis displays the -log10(P) values, indicating the strength of the association. Each dot represents a single SNP, positioned based on its genomic location and statistical significance. The red dashed line marks the genome-wide significance threshold (P = 5 × 10⁻⁸). The statistical test was two-sided, and the Bonferroni-corrected significance level was applied. b Scatter plot of minor allele frequency (MAF) versus effect size (log-odds ratio) for variant-AF associations. Two gray dashed lines indicate MAFs of 0.001 and 0.01. The loci with an effect of odds ratio > 1.3 were labeled with the gene name. c Distribution of loci identified across GWAS of different ancestries. d Venn diagram of shared and unique loci across ancestries. Two loci near PITX2 and ZFHX3 were identified as shared across European (EUR), East Asian (EAS), African (AFR), and Admixed American (AMR) populations. Source data are provided as a Source Data file.

Fig. 3. Pathways enriched based on AF-associated genes.

a Pathway enrichment in the Reactome database. The x-axis represents the effect size of the pathway’s influence on AF, while the y-axis shows the -log10(P) values, indicating statistical significance. Each dot corresponds to a pathway, with blue dots representing pathways that are significant after Bonferroni correction. b Pathway enrichment in the Gene Ontology (GO) database. The analysis includes pathways categorized under biological processes (BP), molecular functions (MF), and cellular components (CC). The x-axis represents the ratio of AF-associated genes to the total number of genes in each pathway, while the y-axis lists the pathways. Each dot represents a pathway, where the color reflects the Bonferroni-adjusted p-value, and the size indicates the count of AF-associated genes in each pathway. For clarity, the figure only highlights the top 10 out of 50 BP pathways due to space constraints. Full results, including all pathways, are provided in Supplementary Data 5 and Supplementary Data 6. The statistical test was two-sided, and the Bonferroni-corrected significance level was applied. Source data are provided as a Source Data file.

Fig. 4. Heterogeneity between Europeans and Africans regarding genetic correlations between atrial fibrillation and other circulatory endpoints.

The analysis was conducted using data from the Million Veteran Program (MVP). The analysis involved 94 correlations both in Europeans and Africans, and heterogeneity was defined by I ² > 75% and P-value for Cochran’s Q < 0.05. The statistical test was two-sided. Detailed information on these genetic correlations is available Supplementary Data 7 and 8. Source data are provided as a Source Data file.

Fig. 5. Genetically predicted associations between 37 modifiable traits and atrial fibrillation (AF).

The estimates and p-values were derived using the inverse variance weighted (IVW) method with a fixed-effects model for traits with ≤ 4 genetic instruments. For traits with > 4 genetic instruments, the results were obtained from MR-PRESSO, accounting for potential pleiotropic effects by removing outlier SNPs where applicable. Detailed results are presented in Supplementary Data 9. Supplementary Data 18 lists the number of instrumental variables, the sample sizes of the source studies, and the units for each trait. The x-axis represents the odds ratio (OR) of AF per unit increase in the genetically predicted trait. Triangles indicate associations with P < 0.05 after Bonferroni correction, while red and blue dots represent positive and inverse associations, respectively. Data are presented as ORs +/− 95% confidence intervals. The statistical test was two-sided, and the Bonferroni-corrected significance level was applied. Source data are provided as a Source Data file.

Fig. 6. Genetically predicted levels of 2847 proteins associated with atrial fibrillation (AF).

We analyzed 2847 unique proteins with cis-instrumental variables derived from the deCODE and UKB-PPP datasets. For proteins present in both datasets, data from UKB-PPP were prioritized due to its larger sample size. All associations were scaled to a one standard deviation increase in genetically predicted protein levels. a volcano plot of protein-AF associations using SMR analysis. The x-axis represents the effect size of protein-AF associations, while the y-axis shows the -log10(P) values. The statistical test was two-sided, and the Bonferroni-corrected significance level was applied. Associations with P < 0.05 after Bonferroni correction and HEIDI test P > 0.05 are labeled. Red and blue dots indicate positive and inverse associations, respectively. b traditional colocalization analysis results. Only protein-AF associations with PPH4 > 0.7 are displayed due to space constraints. The gray line indicates PPH = 0.8, a commonly used threshold for strong colocalization evidence. c SuSiE colocalization analysis results. Similar to panel b, only protein-AF associations with PPH4 > 0.7 are shown. The gray line indicates PPH = 0.8. d forest plot of associations meeting the criteria of Bonferroni-corrected P < 0.05, HEIDI P > 0.05, and colocalization PPH4 > 0.8. Data are presented as ORs +/− 95% confidence intervals. The statistical test was two-sided, and the Bonferroni-corrected significance level was applied. Source data are provided as a Source Data file.

Fig. 7. Polygenic risk score (PGS) and protein score (ProS) for atrial fibrillation (AF) risk prediction.

The analysis for panels (a, b, and c) was based on the Penn Medicine Biobank (PMBB, 4401 individuals with prevalent AF and 32,760 individuals without) and the analysis for panel d was based on the UK Biobank (3441 individuals with incident AF and 47,437 without). Panels (a and b) plots show the prevalence and odds ratio of AF across deciles of our PGS vs. the PGS002814 from the Miyazawa et al. study, respectively. Data in panels (a and b) are presented as mean values +/− SD and ORs +/− 95% confidence intervals, respectively. Panel c plot compares the prediction ability between two PGS (AUC for our PGS = 0.780 and AUC for PGS002814 = 0.767). Panel (d) plot compares the prediction ability between PGS, ProS, and their combination. AUC, area under its receiver operating characteristic curve. Source data are provided as a Source Data file.