Support Vector Regression based QSAR of anti-Haemophilus Influenzae activity of orally administered cephalosporins

Abstract

Support Vector Regression (SVR), a novel robust machine learning technology, was applied to QSAR on the anti-Haemophilus Influenzae (HI) activity of 69 orally active cephalosporins. The optimal model was built with three descriptors-MR, qC7 and qO9, which came from 23 descriptors available. The prediction accuracy of the model was discussed on the basis of Leave-One-Out Cross-Validation (LOOCV) and the independent test dataset. Eighteen newly designed molecules are highly recommended for synthesis scientists based on the SVR model obtained.