Prediction of adverse effects of drug-drug interactions on the cardiovascular system based on the analysis of structure-activity relationships

The co-administration of drugs often leads to drug-drug interactions, which may be accompanied by various adverse drug reactions that pose a threat to the life and health of patients. The effect caused by adverse drug reactions from the cardiovascular system is one of the most significant manifestations of drug-drug interaction. Clinical assessment of adverse drug reactions resulting from drug-drug interaction between all drug pairs used in therapeutic practice is not possible. The purpose of this work was to build models using structure-activity analysis to predict the effects of adverse drug reactions on the cardiovascular system, mediated by interactions between drug pairs when they are taken together. Data on adverse effects resulting from drug-drug interaction were obtained from the DrugBank database. The data on drug pairs that do not cause the corresponding effects, necessary for building accurate structure-activity models, were obtained from the TwoSides database, which contains the results of the analysis of spontaneous reports. Two types of descriptors were used to describe a pair of drug structures: PoSMNA descriptors and probabilistic estimates of the prediction of biological activities obtained using the PASS program. Structure-activity relationships were established using the Random Forest method. Prediction accuracy was calculated by means of a five-fold cross-validation. The highest accuracy values were obtained using PASS probabilistic estimates as descriptors. The area under the ROC curve was 0.94 for bradycardia, 0.96 for tachycardia, 0.90 for arrhythmia, 0.90 for ECG QT prolongation, 0.91 for hypertension, 0.89 for hypotension.

(Q)sar, drug-drug interaction, (Q)sar, adverse drug reactions