Characterizing the genetic architecture of drug response using gene-context interaction methods

Abstract

Identifying the factors responsible for variability of treatment response is a central objective of clinical research. Specially designed pharmacogenomic studies have identified a handful of individual mutations modulating the effect of specific drugs. However, the extent to which drug response variability is driven by genetics is largely unknown, partly due to the small sample sizes of prospective pharmacogenomic trials. In this work, we develop a framework to study the genetic architecture of response to commonly prescribed drugs in large biobanks coupled with electronic health records. Our framework leverages concepts from gene-environment interaction testing, including novel interaction models at the level of genes, polygenic scores (PGS), and genome-wide heritability. We quantified the heritability of response to statins, metformin, warfarin, and methotrexate in 342,257 UK Biobank participants. Our results show that genetic variation modifies the primary effect of statins on LDL cholesterol (9% heritable) as well as its side effects on hemoglobin A1c and blood glucose (10% and 11% heritable, respectively). Next, we identified dozens of specific genes that modify drug response, which we then replicated in a retrospective pharmacogenomic study. Finally, we found that PGS accuracy varies up to 2-fold depending on treatment status, showing that the current approach of building PGS using mostly healthy individuals is likely to underperform in clinical contexts. Together, our results provide a framework for characterizing the genetic architecture of drug response using cross-sectional data.