Abstract: Machine learning methods, including Support Vector Machine, Artificial Neural Network, Regularized Logistic Regression and K-Nearest Neighbor, are used to develop the classification models for a set of 761 DHFR inhibitors. Constitutional descriptors and topological descriptors are calculated to characterize the structural and physicochemical properties of compounds and Kennard-Stone method is used to design the training set and Metropolis Monte Carlo simulated method is used for feature selection. It is shown that SVM method outperforms other machine learning methods used in this study and the final SVM model after feature selection can give a prediction accuracy of 91.62%. This suggests that SVM method with proper training set design and feature selection is potentially useful for the prediction of the activity of a diversity set of DHFR inhibitors.

Key words: Dihydrofolate reductase inhibitor, Support Vector Machine, Molecular descriptor