Sparse kernel logistic regression for β-turns prediction

A β-turn is a secondary protein structure type that plays a significant role in protein folding, stability, and molecular recognition. On average 25% of amino acids in protein structures are located in β-turns. Development of accurate and efficient method for β-turns prediction is very important. Most of the current successful β-turns prediction methods use support vector machines (SVMs) or Neural Networks (NNs), however a method that can yield probabilistic outcome, and has a well-defined extension to the multi-class case will be more valuable in β-turns prediction. Although kernel logistic regression (KLR) is a powerful classification technique that has been applied successfully in many classification problems, however it is often not found in β-turns classification, mainly because it is computationally expensive. In this paper we used KLR to obtain sparse β-turns prediction in short evolution time after speeding it using Nystrom approximation method. Secondary structure information and position specific scoring matrices (PSSMs) are utilized as input features. We achieved Qtotal of 80.4% and MCC of 50% on BT426 dataset. These results show that KLR method with the right algorithm can yield performance equivalent or even better than NNs and SVMs in β-turns prediction. In addition KLR yields probabilistic outcome and has a well-defined extension to multi-class case.