Vladimir Poroikov » Публикация

Аннотация Experimentally found gene expression profiles are used to solve different problems in зharmaceutical studies, such as drug repositioning, resistance, toxicity and drug–drug interactions. A special web service, DIGEP-Pred, for prediction of drug-induced changes of gene expression profiles based on structural formulae of chemicals has been developed. Structure–activity relationships for prediction of drug-induced gene expression profiles were determined by Prediction of Activity Spectra for Substances (PASS) software. Comparative Toxicogenomics Database with data on the known drug-induced gene expression profiles of chemicals was used to create mRNA and protein-based training sets. An average prediction accuracy for the training sets (ROC AUC) calculated by leave-one-out cross-validation on the basis ofmRNA data (1385 compounds, 952 genes, 500 up and 475 down-regulations) and protein data (1451 compounds, 139 genes, 93 up- and 55 down-regulations) exceeded 0.85.