A Novel Algorithm for Hub Protein Identification in H.Sapiens Using Global Amino Acid Features

Abstract

AbstractIdentification of hub proteins solely from amino acids in proteome remains an open problem in computational biology that has been getting increasing deliberations with extensive growth in sequence information. In this context, we have chosen to investigate whether hub proteins can be predicted from amino acid sequence information alone. Here, we propose a novel hub identifying algorithm which relies on the use of conformational, physiochemical and pattern characteristics of amino acid sequences. In order to extract the most potential features, two feature selection techniques, CFS (Correlation-based Feature Selection) and ReliefF algorithms were used, which are widely used in data preprocessing for machine learning problems. The performance of two types of neural network classifiers such as RBF network and multilayer perceptron were evaluated with these filtering approaches. Our proposed model led to successful prediction of hub proteins from amino acid sequences alone with 92.98% and 92.61% accuracy for multilayer perceptron and RBF Network respectively with CFS algorithm and 94.69% and 90.89% accuracy for multilayer perceptron and RBF Network respectively using ReliefF algorithm.KeywordsProtein hubnessProtein protein interaction networksProtein protein interactionfeature selection methodsmachine learning