Prediction of heat shock proteins in plants based on amino acid composition and machine learning methods

Abstract

Heat shock proteins (HSPs) are an important class of proteins which are expressed in cells during extreme biotic or abiotic stress conditions. Rapid identification of the HSPs is crucial in studies related to inducing plant tolerance to abiotic stresses using biotechnological approaches. In the present study we have presented a discrete model based on features of protein sequences namely sequence length along with (i) amino acid compositions (ii) di-peptide compositions and (iii) in combination and machine learning based classifiers viz. decision trees, nearest neighbour and Naïve Bayes for the identification of the heat shock proteins. A classifier for the classification of each class of heat shock proteins (HSP70, HSP90, HSP100 and sHSP) from the remaining sequences has been able developed. Based on the AUC measure, the Naïve Bayes algorithm has been found to be superior in identifying the heat shock proteins in all the classes.