Non-Linear Modeling and Chemical Interpretation with Aid of Support Vector Machine and Regression

Abstract

In quantitative structure-activity relationship (QSAR) and quantitative structure-property relationship (QSPR), there is a considerable interest in support vector machine (SVM) and support vector regression (SVR) for data modeling. SVM and SVR have a high performance for classification and regression rates, but their chemical interpretations are not feasible. In this review, we present some promising approaches to visualize and interpret the SVM and SVR models. This type analysis would be useful for molecular design. Representative examples derived from chemoinformatics and bioinformatics are highlighted in detail. We also refer to a structure generator based on SVR score in the framework of de novo design. Furthermore, we provide readers the theoretical description of SVM and SVR.