80. ANALYSIS OF MISSENSE VARIANT CLUSTERING IMPLICATES NOVEL PROTEIN INTERACTION IN BIPOLAR DISORDER

Background Rare variant association studies have identified genes contributing to psychiatric disease risk, but gene-level associations alone provide limited insight into the molecular mechanisms through which rare coding variation disrupts biology. Missense variants offer a complementary opportunity because they can be mapped onto three-dimensional protein structures, allowing disease-associated variation to be localized to specific functional regions. Bipolar disorder has historically lacked sufficiently powered sequencing datasets for this type of analysis, limiting the ability to move from risk gene discovery to mechanistic interpretation. Methods We analyzed one of the largest global bipolar disorder sequencing datasets to date, comprising over 64,000 bipolar disorder cases and 168,000 controls. Gene-level rare variant association analyses were used to identify bipolar disorder risk genes across ultra-rare protein-truncating and damaging missense variants. To extend these findings beyond gene-level signals, we leveraged a scalable framework for mapping case and control missense variants onto predicted protein structures and testing whether case variants show non-random spatial clustering within proteins. This approach was applied to the top significant bipolar disorder associated genes, with focused evaluation of associated genes and protein regions with evidence of localized case-enrichment. Results Rare variant association analyses identified 13 Bonferroni exome-wide significant bipolar disorder risk genes. Structural mapping of ultra-rare missense variation revealed that disease-associated variants were not uniformly distributed across all protein regions. In DOP1A and ATP9A, case variants showed spatial clustering in localized regions of the predicted protein structures, including regions potentially relevant to protein interaction or functional activity. These findings suggest that rare missense variation in bipolar disorder may act not only through broad gene-level disruption, but also through perturbation of specific structural domains or interaction interfaces. Discussion By integrating large-scale rare variant association data with protein structure predictions, this work provides a framework for moving beyond gene discovery toward mechanistic localization of disease-associated variation. In bipolar disorder, structural clustering in DOP1A and ATP9A highlights candidate functional regions that may be important for cellular trafficking, protein interaction, and neuronal biology. More broadly, these results demonstrate that protein structure-informed RVAS can reveal disease-relevant biology that is not apparent from standard gene-based analyses alone, offering a scalable strategy for functional interpretation of rare coding variation in psychiatric genetics.