Evaluating Drug-Drug Interactions by Constructing Advanced Machine Learning Models of ABC Transporter Inhibitors

ATP-binding cassette (ABC) transporters play a critical role in transmembrane transport, and the inhibition of ABC transporter-mediated efflux can lead to unintended drug-drug interactions (DDIs). The identification of new drug candidates as perpetrators of transporter-mediated DDIs requires synthesis of significant drug products and costly experimental evaluation. Alternatively, computational approaches can rapidly and accurately screen the DDI liability of new drug candidates in silico. Here, we developed a novel DDI risk assessment protocol incorporating classic and graph-based quantitative structure-activity relationship (QSAR) models. These models were trained on publicly available, large-scale, structurally diverse in vitro chemical inhibition data for P-gp, BCRP, and MRP1. The resultant models yielded promising R2 scores (0.58 to 0.82). External predictions successfully identified inhibitors, verified by experimental validation or literature review, and detected drugs with the potential to perpetrate clinically relevant DDIs. In summary, we developed a large, diverse database of ABC transporter inhibitors and present an applicable alternative protocol that incorporates model predictions to evaluate transporter-mediated DDIs for new drug candidates to prevent potentially harmful clinical outcomes.