Application of locality sensitive discriminant analysis to predict protein fold pattern

Abstract

Predicting protein-folding patterns is a challenge due to the complex structure of proteins. Many sequence encoding schemes have been proposed to extract the features of pro-tein sequences, and these features are often fused to form a new combined feature set so that it can contain various useful information. However, there usually has redundant information in the combined features. In this paper, a novel approach, LSDA-SVM, is proposed to predict pro-tein fold pattern. Firstly, protein samples are represented by the pseudo amino acid composition (PseAAC), pair wise feature (PF) and the others five types of protein sequence information, and these features are further combined to form a new feature set. Secondly, the locality sensitive discriminant analysis (LSDA) is employed to extract the more discriminant features. Finally, the support vector machine (SVM) is employed to classify the protein sequences. Experimental results demonstrate the effectiveness of the proposed algorithm.