Interpretation of ANN-based QSAR models for prediction of antioxidant activity of flavonoids.

摘要: Quantitative structure-activity relationships (QSARs) built using machine learning methods, such as artificial neural networks (ANNs) are powerful in prediction of (antioxidant) activity from quantum mechanical (QM) parameters describing the molecular structure, but usually not interpretable. This obvious difficulty is one most common obstacles application ANN-based QSAR models for design potent antioxidants or elucidating underlying mechanism. Interpreting resulting often omitted performed erroneously altogether. In this work, a comprehensive comparative study six methods (PaD, PaD2 , weights, stepwise, perturbation and profile) exploration interpretation ANN Trolox-equivalent antioxidant capacity (TEAC) QM descriptors, presented. Sum ranking differences (SRD) was used with respect to contributions calculated descriptors toward TEAC. The results show that PaD, profile stable give rise realistic observed correlations. Therefore, they safely applicable future interpretations without opinion an experienced chemist bio-analyst. © 2018 Wiley Periodicals, Inc.