AbstractDrugs targeting genes that harbor natural variations associated with the disease the drug is in-dicated for have increased odds to be approved. Various approaches have been proposed to iden-tify likely causal genes for complex diseases, including gene-based genome-wide association stud-ies (GWAS), rare variant burden tests in whole exome sequencing studies (Exome) or integration of GWAS with expression/protein quantitative trait loci (eQTL-GWAS/pQTL-GWAS). Here, we compare gene-prioritization approaches on 30 common clinical traits and benchmarked their ability to recover drug target genes defined using a combination of five drug databases. Across all traits, the top pri-oritized genes were enriched for drug targets with odds ratios (ORs) of 2.17, 2.04, 1.81 and 1.31 for the GWAS, eQTL-GWAS, Exome and pQTL-GWAS methods, respectively. We quantified the perfor-mance of these methods using the area under the receiver operating characteristic curve as metric, and adjusted for differences in testable genes and data origins. GWAS performed significantly better (54.3%) than eQTL (52.8%) and pQTL-GWAS (51.3%), but not significantly so against the Exome ap-proach (51.7%vs52.8% for GWAS restricted to UK Biobank data). Furthermore, our analysis showed increased performance when diffusing gene scores on gene networks. However, substantial improve-ments in the protein-protein interaction network may be due to circularity in the data generation process, leading to the node (gene) degree being the best predictor for drug target genes (OR = 8.7, 95% CI = 7.3-10.4) and warranting caution when applying this strategy. In conclusion, we systematically as-sessed strategies to prioritize drug target genes highlighting promises and potential pitfalls of current approaches.