P-glycoprotein (P-gp), a member of ABC transporter family, actively pumps xenobiotics out of cells to protect organisms from toxic compounds. For a therapeutic drug acting as P-gp substrate, its absorption is hindered and excretion is promoted by P-gp. On the other hand, a P-gp inhibitor can reduce this effect and cause drug-drug interaction (DDI) with P-gp substrates. Hence, it is crucial to know if a drug exhibit P-gp inhibitor or substrate properties. In this work, we introduce PgpRules, an online server with decision tree based P-gp inhibitor and substrate prediction models. The P-gp inhibitor and substrate predicting models were constructed from 2056 and 925 compounds, respectively. Both of the models were built using classification and regression tree (CART) algorithm, which generates binary decision trees by maximizing the predictive accuracy when making every new branching. Three descriptor sets including PubChem 2D Fingerprint, RDKit Descriptors and composite of both were tested. The classification accuracy of the best P-gp inhibitor and substrate model was 0.86 and 0.70, respectively. The branching rules also offer specific chemical feature guidance to refine compounds toward the requested P-gp properties. Thus, within acceptable execution time, PgpRules can contribute to <i>in silico</i> drug design in virtual high-throughput screening (VHTS).

PgpRules: A decision tree based P-glycoprotein inhibitor and substrate prediction server

2019 Spring National Meeting

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.