Quantitative structure activity relationship study on IC50 of novel evodiamine derivatives against gastric cancer by machine learning

Zixuan Yang, Guoliang Li

Abstract: A quantitative structure activity relationship (QSAR) study was performed based on a data set of 42 novel evodiamine derivatives that can against stomach cancer. 551 chemical descriptors were calculated from molecular structure of the compounds. The QSAR models were built using the heuristic method (HM), support vector machine (SVM) and gene expression programming (GEP). Model established by SVM is the best one among all these three models generated, with R^2=0.9542, for training set and R^2=0.9316 for test set. The leave-one-out (LOO) method was chosen as the cross-validation method, and the result of LOO is R^2=0.8924. The predicted results of SVM model are consistent with the experimental results. This study reveals 4 key descriptors of novel evodiamine derivatives and will help to screen out novel and effective drugs in the future.

: Download (free)  CSE-PDF

Inhibitory prediction of thiazole derivatives based on machine learning for MCF-7

Tianci Huang, Jiajian Huang, Yili Wang, Peijian Zhang

Abstract: The quantitative structure activity relationship (QSAR) study was carried out to predict the anti-breast cancer effect of 2-Aminothiazole derivatives. heuristic method (HM), random forest (RF), support vector machine (SVM), and genetic expression programing (GEP), four QSAR models were established based on four descriptors. And the model built by SVM showed the best prediction ability and strongest model robustness: In the training set, R2 = 0.959, RMSE = 0.010. In the test set, R2=0.941 RMSE=0.013. This experimental study showed the role of the four key descriptors of 2-Aminothiazole derivatives in the treatment of breast cancer, which can provide a new basis for the prediction and screening of new drugs.

: Download (free)  CSE-PDF