QSAR study of human epidermal growth factor receptor (EGFR) inhibitors: conformation-independent models

Abstract

Many compounds have been proposed and tested as human epidermal growth factor receptor (EGFR) inhibitors for cancer treatment. Recently, new survival mechanisms of cancer cells have been discovered with the consequent resistance to therapy, which makes it necessary to search for new anticancer drugs. Here we perform a quantitative structure-activity relationship (QSAR) study on 290 compounds reported in the literature as EGFR inhibitors to analyze the molecular properties that may influence their activity. A large number of nonconformational descriptors (17,974) were explored including molecular descriptors, flexible molecular descriptors, and combination of both. To avoid ambiguities derived from the existence of several conformational states, only constitutional and topological molecular descriptors have been considered. The models were validated through Y-randomization, cross-validation, and mean absolute error criteria. A simple model involving flexible descriptors shows the best predictive performance and suggests that the presence of multiple aromatic rings and amino groups in a compound structure may increase its EGFR inhibitory activity.