Drug-Drug Interactions prediction from enzyme action crossing through machine learning approaches

Drug-Drug Interactions (DDIs) are major causes of morbidity and treatment inefficacy. The prediction of DDIs for avoiding the adverse effects is an important issue. There are many drug-drug interaction pairs, it is impossible to do in vitro or in vivo experiments for all the possible pairs. The limitation of DDIs research is the high costs. Many drug interactions are due to alterations in drug metabolism by enzymes. The most common among these enzymes are cytochrome P450 enzymes (CYP450). Drugs can be substrate, inhibitor or inducer of CYP450 which will affect metabolite of other drugs. This paper proposes enzyme action crossing attribute creation for DDIs prediction. Machine learning techniques, k-Nearest Neighbor (k-NN), Neural Networks (NNs), and Support Vector Machine (SVM) were used to find DDIs for simvastatin based on enzyme action crossing. SVM preformed the best providing the predictions at the accuracy of 70.40 % and of 81.85 % with balance and unbalance class label datasets respectively. Enzyme action crossing method provided the new attribute that can be used to predict drug-drug interactions.