Discovery of novel EGFR and BRAF inhibitors by machine learning approach

In the pharmaceutical industry, epidermal growth factor receptor (EGFR) and B-Raf proto-oncogene, serine/threonine kinase (BRAF) are the primary targets for targeted cancer therapy. This study aimed to develop novel models for predicting the bioactivity of EGFR and BRAF inhibitors and to uncover critical substructural groups for EGFR and BRAF targets. K-nearest neighbor, Gaussian naive Bayes, random forest (RF), decision tree, and support vector machine models were utilized to develop classifiers for EGFR and BRAF inhibitors. In addition, the critical functional groups of these inhibitors were analyzed, and molecular docking was performed. The RF model was found to be the preferred choice for predicting the bioactivity of EGFR and BRAF inhibitors. The RF model for BRAF had an accuracy and area under the curve (AUC) of 93% and 0.96, respectively, and the RF model for EGFR had an accuracy and AUC of 90% and 0.96, respectively. The results of this study show the importance of properly increasing chlorine or fluorine in inhibitors of BRAF or EGFR. Moreover, the EGFR and BRAF inhibitors exhibited the same halogen selection priority of fluorine > chlorine > bromine. The present study develops novel models for predicting the bioactivity of EGFR and BRAF inhibitors. The hydrogen bond formation is a key factor for influencing the selection priority of fluorine > chlorine > bromine.