Thinking Beyond Association Tests: Using Mathematical Models to Identify Genes Influencing Exceptional Longevity

The main goal in the genetic study of trait biology is to identify genetic factors, such as variants and genes, that causally influence traits and to understand the downstream mechanisms through which they exert their effects. This dissertation improves our understanding of trait biology by building on widely applied omics-wide association studies (omics-WAS), such as GWAS and TWAS, and proposing methods that (1) combine association statistics from multiple omics modalities and (2) systematically incorporate prior knowledge from biological networks and functional annotations with omics-WAS summary statistics. First, we apply meta-analysis to integrate gene-level p-values from GWAS, TWAS, and rare-variant analyses to identify novel and replicable genes associated with cardiovascular risk in the Long Life Family Study (LLFS) cohort. This approach improves statistical power, reduces confounding effects in individual omics-WAS, and prioritizes gene-trait relationships supported by multiple omics-WAS. Second, we develop FISHNET (FInding Significant Hits in NETworks), which incorporates gene interaction networks and functional annotations to identify genes that do not meet genome-wide significance thresholds but replicate nonetheless. Third, we reconstruct transcription factor (TF) network maps to identify direct and functional TF-target gene relationships in a tissue-specific context. By integrating high- throughput omics datasets that capture different aspects of TF-target gene regulation, these network maps provide insights into TF-TG relationships that may mediate variant/gene-trait associations in omics-WAS. Together, these three approaches advance the interpretation of statistical associations from individual omics-WAS, contributing to a deeper understanding of trait biology.