When several drugs are co-administrated, a drug-drug interaction (DDI) phenomenon may appear. Many DDIs are due to changes in the metabolism of drugs. In this case, DDI is manifested by the effect of one drug on the biotransformation of other drugs, its slowdown (in case of inhibition of drug-metabolizing enzymes) or acceleration (in case of induction of drug-metabolizing enzymes), which leads to change in the pharmacological action of co-administrated drugs.<br/>The severity of DDIs can be classified by applying different classification systems. One of the most advanced is OpeRational ClassificAtion (ORCA) system for the classification of DDI, created for physicians to assess the risk of co-administration of two drugs. ORCA divides DDI into five classes: contraindicated (class 1), provisionally contraindicated (class 2), conditional (class 3), minimal risk (class 4), no interaction (class 5). For computer prediction of the severity of DDI, we collected a training set consisting of approximately 4000 pairs from 500 drugs that belong to the ORCA classes 1-3 of DDI in case of co-administration. The prediction of DDI classes is based on a combination of substructural MNA descriptors (Multilevel Neighbourhoods of Atoms), and naïve Bayesian classification algorithm implemented in the PASS (Prediction of Activity Spectra for Substances) software.<br/>The average invariant accuracy of prediction, calculated in the leave-one-out and 20-fold cross-validation procedures, were approximately 0.9.<br/>The drug-drug interactions severity prediction based on xenobiotic structural formulas and the PASS prediction algorithm provides accuracy sufficient for practical application in the fields of the biomedical and pharmaceutical chemistry. Such prediction prior to clinical use can increase the safety of new drugs being developed and their application in medical practice.

Drug-drug interactions severity prediction using PASS approach

256th National Meeting (2018)

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