Sequentially distant but structurally similar proteins exhibit fold specific patterns based on their biophysical properties. (August 2018)

• Record Type: Journal Article
• Authors: Rajendran, Senthilnathan; Jothi, Arunachalam
• Is Part Of: Computational biology and chemistry. Volume 75(2018)
• Page Start: 143
• Page End: 153
• Publication Date: 2018-08
• Subjects: CATH classification -- Biophysical descriptors -- Naive Bayes -- SVM -- BGLM -- PCA -- Amino acid properties; 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.2018.05.009
• Languages: English
• ISSNs: 1476-9271

Abstract

Graphical abstract: Highlights: Proteins can share a same structural fold but insignificant sequence similarity. Sequentially distant proteins can be discriminated based on their biophysical properties. Biophysical properties show a fold specific correlation patterns. Binary classifiers can classify structural classes with upto 83% accuracy. Abstract: The Three-dimensional structure of a protein depends on the interaction between their amino acid residues. These interactions are in turn influenced by various biophysical properties of the amino acids. There are several examples of proteins that share the same fold but are very dissimilar at the sequence level. For proteins to share a common fold some crucial interactions should be maintained despite insignificant sequence similarity. Since the interactions are because of the biophysical properties of the amino acids, we should be able to detect descriptive patterns for folds at such a property level. In this line, the main focus of our research is to analyze such proteins and to characterize them in terms of their biophysical properties. Protein structures with sequence similarity lesser than 40% were selected for ten different subfolds from three different mainfolds (according to CATH classification) and were used for this analysis. We used the normalized values of the 49 physio-chemical, energetic and conformational properties of amino acids. We characterize the folds based on the average biophysical property values. We also observed a fold specific correlational behavior of biophysical properties despite a very low sequence similarity in our data. We further trained three different binary classification models (Naive Bayes-NB, Support Vector Machines-SVM and Bayesian Generalized Linear Model-BGLM) which could discriminate mainfold based on the biophysical properties. We also show that among the three generated models, the BGLM classifier model was able to discriminate protein sequences coming under all beta category with 81.43% accuracy and all alpha, alpha-beta proteins with 83.37% accuracy.