Bayes PoPS:A Novel Bayesian Approach for Improved Polygenic Gene Prioritization

<b>Background</b>: Polygenic prioritization scoring (PoPS) integrates genome-wide association study (GWAS) summary statistics to identify genes associated with complex traits and diseases. It combines polygenic signals, linkage disequilibrium, and functional annotations, to assign a prioritization score to each gene. We demonstrate superior performance of a novel extension, namely Bayes PoPS.<br/><b>Material and Methods</b>: Bayes PoPS incorporates BayesR prior, eliminating the initial feature selection in PoPS. We simulated quantitative phenotypes using genotypic data from 335,000 UK Biobank participants across eight different genomic scenarios (Table), with ten independent replicates for each scenario. To assess various gene-set characteristics, we generated sets containing 20 to 100 genes, with five or twenty designated as causal (nc). We calculated VEGAS gene-level association statistics, and derived PoPS and Bayes PoPS scores by utilising the gene-sets. The performance of each gene prioritization method was evaluated by area under the receiver operating characteristic curve (AUC).<br/><b>Results</b>: Across all simulations, Bayes PoPS consistently outperformed PoPS (P &lt; 2.2∙10-16). In scenarios with a higher proportion of genome-wide causal SNPs and few causal genes within gene-sets (nc=5), VEGAS outperformed both PoPS and Bayes PoPS. The power of Bayes PoPS to correctly prioritize genes significantly improved with a higher fraction of causal genes within gene-sets (nc=20).<br/><b>Conclusion</b>: Bayes PoPS demonstrated improved power to prioritize genes and robustly distinguish causal genes from polygenic background. This enhances gene discovery and interpretation in complex trait genetics, offer insights for downstream functional analyses and personalized medicine applications.<br/>