关键词: 残基远程相互作用预测, 蛋白质三级结构预测, 图模型, 共进化, 机器学习

Abstract: Proteins are large molecules consisting of a linear sequence of amino acids. In the natural environment, a protein spontaneously folds into specific tertiary structure to perform its biological functionality. The main factors that drive proteins to fold are interactions between residues, including hydrophobic interaction, Van der Waals’ force and electrostatic interactions. The interactions between residues usually lead to residue-residue contacts, and the prediction of residue-residue contacts should greatly facilitate understanding of protein structures and functionalities. A great variety of techniques have been proposed for residue-residue contacts prediction, including machine learning, statistical models, and linear programing. It should be pointed out that most of these techniques are based on the biological insight of co-evolution, i.e., during the evolutionary history of proteins, a residue’s mutation usually leads its contacting partner to mutate accordingly. In this review, we summarize the state-of-art algorithms in this field with emphasis on the construction of statistical models based on biological insights. We also present the evaluation of these algorithms using CASP (critical assessment of techniques for protein structure prediction) targets as well as popular benchmark datasets, and describe the trends in the field of protein contact prediction.

Key words: protein contact prediction, protein tertiary structure prediction, graphical model, co-evolution, machine learning