Identifying Native and Non-native Membrane Protein Loops by Using Stabilizing Energetic Terms of Three Popular Force Fields

Abstract

Background: Predicting the three-dimensional structure of globular proteins from their amino acid sequence has reached a fair accuracy, but predicting the structure of membrane proteins, especially loop regions, is still a difficult task in structural bioinformatics. The difficulty in predicting membrane loops is due to various factors like length variation, position, flexibility, and they are easily prone to mutation. Objective: In the present work, we address the problem of identifying and ranking near-native loops from a set of decoys generated by Monte-Carlo simulations. Methods: We systematically analyzed native and generated non-native decoys to develop a scoring function. The scoring function uses four important stabilizing energy terms from three popular force fields, such as FOLDX, OPLS, and AMBER, to identify and rank near-native membrane loops. Results: The results reveal better discrimination of native and non-natives and perform poor prediction in binary classifying native and near-native defined based on Root Mean Square Deviation (RMSD), Global Distance Test (GDT), and Template Modeling (TM) score, respectively. Conclusions: From our observations, we conclude that the important energy features described here may help to improve the loop prediction when the membrane protein database size increases.