Using factor analysis scales of generalized amino acid information for prediction and characteristic analysis of β-turns in proteins based on a support vector machine model

Abstract

This paper offers a new combined approach to predict and characterize β-turns in proteins. The approach includes two key steps, i.e., how to represent the features of β-turns and how to develop a predictor. The first step is to use factor analysis scales of generalized amino acid information (FASGAI), involving hydrophobicity, alpha and turn propensities, bulky properties, compositional characteristics, local flexibility and electronic properties, to represent the features of β-turns in proteins. The second step is to construct a support vector machine (SVM) predictor of β-turns based on 426 training proteins by a sevenfold cross validation test. The SVM predictor thus predicted β-turns on 547 and 823 proteins by an external validation test, separately. Our results are compared with the previously best known β-turn prediction methods and are shown to give comparative performance. Most significantly, the SVM model provides some information related to β-turn residues in proteins. The results demonstrate that the present combination approach may be used in the prediction of protein structures.