Similarity/dissimilarity analysis of protein structures based on Markov random fields. (August 2018)

• Record Type: Journal Article
• Authors: Wu, Jiaqi; Zhou, Tao; Tao, Jin; Hai, Yabing; Ye, Fei; Liu, Xiaoqing; Dai, Qi
• Is Part Of: Computational biology and chemistry. Volume 75(2018)
• Page Start: 45
• Page End: 53
• Publication Date: 2018-08
• Subjects: Protein structures -- Similarity/dissimilarity analysis -- Contact map matrix -- Markov random fields -- Cyclic peptides; Chemistry -- Data processing -- Periodicals; Biology -- Data processing -- Periodicals; Biochemistry -- Data processing; Biology -- Data processing; Molecular biology -- Data processing; Periodicals; Electronic journals; 542.85
• DOI: 10.1016/j.compbiolchem.2018.04.016
• Languages: English
• ISSNs: 1476-9271

Abstract

Graphical abstract: The 2-point cliques, 4-point cliques and 8-point cliques of Markov random fields. Highlights: We first introduced Markov random fields to compare the protein structures. We defined a distance matrix based on different atom sets at residue fragment level. We revised the contact map matrices with a standard statistical distribution. We discussed the influence of the different cliques of Markov random fields. Abstract: Protein Structure Similarity plays an important role in study on functional properties of proteins and evolutionary study. Many efficient methods have been proposed to advance protein structural comparison, but there are still some challenges in the contact strength definitions and similarity measures. In this work, we schemed out a new method to analyze the similarity/dissimilarity of the protein structures based on Markov random fields. We evaluated the proposed method with two experiments and compared it with the competing methods The results indicate that the proposed method exhibits a strong ability to detect the similarities/dissimilarities among the conformation of different cyclic peptides and protein structures. We also found that the alpha-C, oxygen O and N allow us to extract more conserved structures of the proteins, and Markov random fields with 2-point cliques (V) and orders 3 and 1 are more efficient in detecting the similarities/dissimilarities among different protein structures. This understanding can be used to design more powerful methods for similarities/dissimilarities analysis of different protein structures.