Predicting AXL Inhibition of Chemicals using Molecular Descriptors and Machine Learning Methods

Breast cancer has become an emerging medical concern for women because of its intrusive potential of cynical growth factors and progressive rate of metastasis leading up to mortality with an alarming incidence rate so far. And lack of proper treatment or effective drugs can result in the poor survival of the patients affected. Since EMT endowing cancer cells work as a critical driver of metastasis, its effectors are supposed to serve prognostic sapience in establishing an affluent drug discovering strategy following AXL being a dominant factor in encountering the severity of the metastasis stage. AXL has been a conversant target showing prompt sensitivity for GAS6 signaling that is subsequently the predominant effector for cancer cell proliferation, metastasis, and tumor-promoting mechanisms. Our research work presented here can provide a technique so that the drug discovery of AXL inhibitors can be more effective and less time-consuming. The numerical value of bioactivity unit IC50 is considered the critical parameter for declaring an organic compound as active and inactive for AXL inhibition. Our thorough study has conducted on a dataset where 527 compounds were used to predict the value of bioactivity for AXL inhibition. Four different molecular descriptors deliberated in Lipinski’s rule of five that were calculated to build and train eight different classification models to enumerate the bioactivity values where the models were further evaluated using ten-fold cross-validation. Among all the classification models, Gradient Boosting has emerged to account the best performance with an accuracy of 85% and a F1-score of 0.77.