79. GENETIC DIVERSITY REFINES THE ARCHITECTURE OF BIPOLAR DISORDER RISK ACROSS HUMAN POPULATIONS

Background Bipolar disorder is highly heritable but has an extreme polygenic basis, with hundreds of identified loci exerting small but replicable effects, and thousands more contributing to aggregate liability despite going undetected at established genome-wide significance thresholds. To date, bipolar disorder's genetic architecture has been defined largely in populations of European ancestry, constraining biological insight. Methods Extending our recent work in ancestrally diverse genome-wide association studies (GWAS), we leverage electronic health record–linked genomic data and global cohorts to perform the largest multi-ancestry genome-wide association study of bipolar disorder to date, with substantial expansion of African (AFR) and admixed American (AMR) populations, including individuals from the Million Veteran Program (MVP), Cooperative Studies Program (CSP), All of Us (AOU) Research Program, and Genomic Psychiatry Cohort (GPC) studies, totaling 12,567 AFR-like cases and 62,198 controls and 5,871 AMR-like cases and 45,445 controls. Combining our results with available summary statistics from 23andMe, Inc increased these sample sizes to 16,493 AFR-like and 17,861 AMR-like affected persons, and 459,206 controls. Subsequent meta-analysis with European (EUR) and East Asian (EAS) data from CSP/MVP, FinnGen, GPC, the Psychiatric Genomics Consortium (PGC), and 23andMe, Inc culminated in a trans-ancestry GWAS of 201,610 affected and 3,462,410 unaffected persons. To improve our understanding of underlying biological mechanisms shared across populations, we integrated fine-mapped credible causal sets with single-cell functional genomic data from the human brain. Results We identify novel ancestry-specific risk loci, including EML6–RTN4, DLGAP2, and ZNF613 in African-ancestry populations, and ZNF423 in admixed American cohorts, with local ancestry analyses further implicating a tract-specific signal at RORA. Multi-ancestry meta-analysis across European, African, admixed American, and East Asian populations (112,441 cases and 1.3 million controls) identifies 192 independent loci, increasing to 400 with inclusion of self-reported data (202,529 cases and 3.26 million controls). Leveraging differences in linkage disequilibrium across populations, we develop BLENDED-LD (BL50), an ancestry-aware fine-mapping framework that improves model convergence and refines credible sets. Integration with brain transcriptomic and regulatory datasets reveals convergent signals across single-nucleus transcriptome-wide association studies, enhancer–promoter interaction maps, and chromatin accessibility profiles, prioritizing 63 genes and implicating cell-type-specific regulatory mechanisms in GABAergic neurons and oligodendrocytes. Discussion Together, these results show that genetic diversity enhances not only discovery but causal resolution, revealing a more precise and biologically grounded architecture of bipolar disorder risk. Our developed methods generate stable, interpretable credible sets and prioritized candidate variants for downstream mechanistic follow-up studies (e.g. CRISPR perturbation).