For 2D-QSAR study, the data set containing inhibitors and noninhibitors was randomly divided into three training sets which accounted for 75%, 70%, and 50% of the whole data set, respectively (see Supplementary Material 1, available online at https://doi.org/10.1155/2017/4649191).
In this study, tenfold cross-validation test and independent set test were applied to evaluate the prediction ability of the 2D-QSAR model.
Feature Selection and the 2D-QSAR Prediction Model.
Based on the optimal features subset, the SVM classifier method was used to build the 2D-QSAR prediction model.
Sharma showed the 2D-QSAR studies of cSrc tyrosine kinase inhibitors with [q.sup.2] = 0.755 and [r.sup.2] = 0.832 .
In this study, 2D-QSAR and 3D-QSAR prediction models were built to analyze EGFR inhibitors.