Protein Fold Recognition Using Genetic Algorithm Optimized Voting Scheme and Profile Bigram

Abstract

In biology, identifying the tertiary structure of a protein helps determine its functions. A step towards tertiary structure identification is predicting a protein's fold. Computational methods have been applied to determine a protein's fold by assembling information from its structural, physicochemical and/or evolutionary properties. It has been shown that evolutionary information helps improve prediction accuracy. In this study, a scheme is proposed that uses the genetic algorithm (GA) to optimize a weighted voting scheme to improve protein fold recognition. This scheme incorporates k-separated bigram transition probabilities for feature extraction, which are based on the Position Specific Scoring Matrix (PSSM). A set of SVM classifiers are used for initial classification, whereupon their predictions are consolidated using the optimized weighted voting scheme. This scheme has been demonstrated on the Ding and Dubchak (DD), Extended Ding and Dubchak (EDD) and Taguchi and Gromhia (TG) datasets benchmarked data sets.