Abstract
Genome-wide association studies of common disorders reveal polygenic architectures. While these variants may converge into biologically relevant pathways, how polygenicity aggregates across pathways remains unknown. We studied polygenicity at the pathway level in anorexia nervosa (AN)-a psychiatric disorder with multi-systemic clinical manifestations. We constructed pathway-based polygenic risk scores (pathway PRS) to model genetic risk at the pathway level for 3,687 individuals with AN and 11,257 controls. We identified 497 AN-associated pathways after Bonferroni correction, including pathways involved in the brain, metabolism, immunity/stress, and reproduction/development. A strong positive correlation was observed between number of pathways ranking at top 10% and AN case proportion (r = 0.74, P = 7.98×10(-21)). The positive correlation between pathway count and AN risk was also observed when restricted to individuals with low overall genetic risk of AN. Pathways ranking at top 10% among cases were more diverse than those in controls. Higher AN risk was observed for pathway aggregation within function (e.g., brain or metabolism) and across functions (e.g., brain + metabolism). Individuals at top 10% risk of brain-brain and brain-metabolism pathway pairs had the highest average AN risk compared to other pathway pairs (e.g., brain-immune, metabolism-metabolism). Further, pathway PRS provided higher prediction power of AN than overall genome-wide PRS and contributed to the genetic liability of AN among AN cases with low overall genetic risk. Together, our results demonstrate that polygenicity exists at the pathway level, opening new avenues for disease prediction and identification of actionable targets.