MULTI-OMICS INTEGRATION IDENTIFIES GENES INFLUENCING TRAITS ASSOCIATED WITH CARDIOVASCULAR RISKS

Abstract

Abstract The Long Life Family Study (LLFS) enrolled 4,953 participants in 539 pedigrees displaying exceptional longevity. To identify genetic mechanisms that protect LLFS participants against age-related cardiovascular risks, we developed a freely available multi-omics integration pipeline and applied it to 11 traits associated with cardiovascular risks. Using our pipeline, we aggregated gene-level statistics from Rare-Variant Analysis, GWAS, and gene expression-trait association by Correlated Meta-Analysis (CMA). Across all traits, CMA identified 51 significant genes after Bonferroni correction (P ≤ 2.8×10-7). CETP, NLRC5, SLC45A3, and TOMM40 lie within 50 Kb of a known trait-associated variant (previously associated genes). Analysis of protein-protein interaction (PPI) networks identified another 63 genes (passing genes) that (1) have CMA p-value ≤ 5×10-3, (2) lie in a PPI module (highly connected subnetwork) enriched for genes with low P-values, and (3) are annotated with a biological process that is enriched among module genes, ten of which were previously associated with the same traits. Permutation analysis showed that passing genes have a false positive rate of 1 in 14876 and are more likely to be previously known than non-passing genes with similar p-values. CMA improved on the 3 input analyses by producing the largest number of modules enriched for genes with low P-values and highly enriched for genes participating in shared biological processes. Overall, module analysis identified highly plausible candidate causal genes whose P-values after CMA alone were merely suggestive.