Pseudo amino acid composition improves antifreeze protein prediction

Abstract

Antifreeze proteins (AFP) in living organisms play a key role in their tolerance to extremely cold temperatures and have wide range of biotechnological applications. But on account of diversity, their identification has been challenging to biologists. Earlier work explored in this area did not cover introduction of sequence order information, known to represent important properties of various proteins and protein systems for prediction of their attributes. In this study, the effect of Chou's pseudo amino acid composition that presents sequence order of proteins was systematically explored using support vector machines for AFP prediction. Our findings suggest that introduction of sequence order information helps identify AFPs with an accuracy of 84.75% on independent test dataset, outperforming approaches such as AFP-Pred and iAFP. The relative performance calculated using Youden's Index (Sensitivity + Specificity -1) was found to be 0.71 for our predictor (AFP-PseAAC), 0.48 for AFP-Pred and 0.05 for iAFP. We hope this novel prediction approach will aid in AFP based research for biotechnological applications.