Comparative classification study of toxicity mechanisms using support vector machines and radial basis function neural networks

• Published in: Analytica chimica acta Vol. 535; no. 1; pp. 259 - 273
• Main Authors: Yao, X.J., Panaye, A., Doucet, J.P., Chen, H.F., Zhang, R.S., Fan, B.T., Liu, M.C., Hu, Z.D.
• Format: Journal Article
• Language: English
• Published: Amsterdam Elsevier B.V 01.04.2005; Elsevier
• Subjects: Analytical chemistry; Chemistry; Classification; Exact sciences and technology; General, instrumentation; Narcosis mechanism; Phenols; Radial basis function neural network; Support vector machine; Toxicity; Phenols; Narcosis mechanism; Support vector machine; Toxicity; Radial basis function neural network; Classification; Performance evaluation; Discriminant analysis; Neural network; Radial basis function; Phenol; Structure activity relation; Comparative study
• ISSN: 0003-2670; 1873-4324
• DOI: 10.1016/j.aca.2004.11.066

The performance and predictive capability of support vector machine (SVM) and radial basis function neural network (RBFNN) for classification problems in QSAR/QSPR were investigated and compared with several other classification methods such as linear discriminant analysis (LDA) and nonlinear discriminate analysis (NLDA). In the present study, two different data sets are evaluated. The first one involves the classification of four action modes of 221 phenols and the second investigation deals with the classification of the three narcosis mechanism of aquatic toxicity for 194 organic compounds. In both cases, the predictive ability of the SVM model is comparable or superior to those obtained by LDA, NLDA and RBFNN. The obtained results indicate that the SVM model with the RBF kernel function can be used as an alternative tool for classification problems in QSAR/QSPR.