An optimized prediction framework to assess the functional impact of pharmacogenetic variants

Prediction of phenotypic consequences of mutations constitutes an important aspect of precision medicine. Current computational tools mostly rely on evolutionary conservation and have been calibrated on variants associated with disease, which poses conceptual problems for assessment of variants in poorly conserved pharmacogenes. Here, we evaluated the performance of 18 current functionality prediction methods leveraging experimental high-quality activity data from 337 variants in genes involved in drug metabolism and transport and found that these models only achieved probabilities of 0.1-50.6% to make informed conclusions. We therefore developed a functionality prediction framework optimized for pharmacogenetic assessments that significantly outperformed current algorithms. Our model achieved 93% for both sensitivity and specificity for both loss-of-function and functionally neutral variants, and we confirmed its superior performance using cross validation analyses. This novel model holds promise to improve the translation of personal genetic information into biological conclusions and pharmacogenetic recommendations, thereby facilitating the implementation of Next-Generation Sequencing data into clinical diagnostics.