Quantitative Structure-Activity Relationship Modeling to Predict Drug-Drug Interactions Between Acetaminophen and Ingredients in Energy Drinks

Abstract

The evaluation of drug-drug interactions (DDI) is a crucial step in pharmaceutical drug discovery and design. Unfortunately, if adverse effects are to occur between the co-administration of two or more drugs, they are often difficult to test for. Traditional methods rely on in vitro studies as a basis for further in vivo assessment which can be a slow and costly process that may not detect all interactions. Here is presented a quantitative structure-activity relationship (QSAR) modeling approach that may be used to screen drugs early in development and bring new, beneficial drugs to market more quickly and at a lesser cost. A data set of 6532 drugs was obtained from DrugBank for which 292 QSAR descriptors were calculated. The multi-label support vector machines (SVM) method was used for classification and the K-means method was used to cluster the data. The model was validated in vitro by exposing Hepa1-6 cells to select compounds found in energy drinks and assessing cell death. Model accuracy was found to be 99%, predicting 50% of known interactions despite being biased to predicting non-interacting drug pairs. Cluster analysis revealed interesting information, although current progress shows that more data is needed to better analyse results, and tools that bring various drug information together would be beneficial. Non-transfected Hepa1-6 cells exposed to acetaminophen, pyridoxine, creatine, L-carnitine, taurine and caffeine did not reveal any significant drug-drug interactions, nor were they predicted by the model.