PharmGScore scores of compound genetic variant burden for psychiatric treatment optimization

The acceptability of antidepressant drugs partly depends on genetic factors. The list of genes involved in antidepressant response, including Adverse Drug Reactions (ADRs) is broad and contains both drug-metabolizing enzymes (pharmacogenes) and genes involved in pharmacodynamics. Variants in pharmacogenes are traditionally reported in the form of star alleles and are partially annotated with known phenotypic consequences. As it is unfeasible to analyze all genotype-phenotype pairs, computational approaches remain the practical solution. A pharmacogenetic framework to predict responses to antidepressant drug treatment would provide great benefit to patients. In this study, we present a scoring system (PharmGScore) to assess both rare and common genetic variant burden across multiple genes. The PharmGScore is constructed by normalizing and aggregating existing, well-established computational variant predictors (CADD, Fathmm-xf, PROVEAN, Mutation Assessor). We show that this score effectively distinguishes no and decreased function from normal and increased function pharmacogenetic variants reported in PharmVar (PharmGScore AUC = 0.86). PharmGScore has improved performance when compared to its component scores (AUCs: CADD = 0.79; FATHMM-XF = 0.81; PROVEAN = 0.81; Mutation Assessor = 0.75). We then apply the PharmGScore to the 200k exome sequences of the UK Biobank (UKB). We report the overrepresentation of UKB participants with high (>50) gene PharmGScore for CYP2C19 and CYP2C9 and with high (>100) compound PharmGScore from nine pharmacogenes within a group with an antidepressant toxicity diagnostic code (T43.2). We then analyze all UKB participants that received any antidepressant toxicity or ADR diagnosis (n = 602). We indicate genes for which a higher burden may be associated with antidepressant toxicity or ADRs and confirm the known roles of CYP2C19 and CYP2D6 in this process. Finally, we show that patients who experienced ADRs to antidepressants in the therapeutic process or accidental poisoning with antidepressants have a higher PharmGScore composed of nine cytochrome P450 genes. Our study proposes a novel paradigm to assess the compound genetic variant burden associated with antidepressant response from exome sequencing data. This approach can be further applied to a user-defined set of genes to investigate other pharmacological traits.