Integration of genetic evidence to identify approved drug targets

Abstract

Drugs targeting genes supported by human genetic evidence are more likely to succeed in clinical trials. While previous approaches have benchmarked individual methods such as genome-wide association studies (GWAS), rare variant burden testing, and quantitative trait locus (QTL)-informed Mendelian randomization, it remains unclear how best to integrate these signals for drug target discovery. Here, we compared gene-prioritization strategies across 30 complex traits, evaluating their ability to recover approved drug targets compiled into lenient and moderate gold-standard sets from six curated databases. Gene-level association scores from GWAS, expression QTL, protein QTL, and exome-based analyses were integrated using five unsupervised approaches. Predictive performance was assessed with area under the receiver operating characteristic curve (AUROC) and enrichment-based statistics. Across traits, GWAS alone ranked known drug targets on average [~]652 ranks (3.42%) above random expectation, and the minimum-rank-based integration strategy provided an improvement of further [~]558 positions (2.93%), achieving the best AUROC in 23 of 30 traits. When comparing genetic correlation and drug target overlap across trait pairs, we observed a significant positive association (r = 0.193; p = 5.46e-5), and cross-trait analyses further revealed that prioritization scores derived from related diseases could at times equal or even surpass a traits own performance. For instance, coronary artery disease data improved the prediction of stroke targets (p = 0.004), while inflammatory bowel disease data enhanced the prioritization of chronic kidney disease targets (p = 0.014). Taken together, these results demonstrate that integrating complementary genetic signals through a minimum-rank-based framework, combined with information from genetically related traits, systematically strengthens drug target identification across complex diseases.