Refinement and improvement of template based protein modelling algorithms

Abstract

Protein structure prediction algorithms are studied to construct accurate models of the newlineprotein sequences for bridging the ever-increasing gap between the available count of protein newlinesequences and the experimentally determined protein structures. Comparative modelling is newlineconsidered as most popular and accurate structure prediction algorithm to model protein newlinestructure. Template selection is considered as one of the most important steps of a newlinecomparative modelling algorithm. However, selection of the best set of templates is still a newlinemajor challenge. An effective template ranking algorithm is developed to efficiently select newlineonly the reliable hits for predicting the accurate protein structures. The algorithm employs the newlinepairwise as well as multiple sequence alignments of template hits to respectively capture their newlinekey sequence and structural information based scores for effectively ranking them, selecting newlinetheir best possible set and constructing an accurate target model. Modelling accuracy of the newlinealgorithm is tested and evaluated on TBM-HA domain containing CASP8, CASP9 and newlineCASP10 targets. In-house C, Python and PERL scripts are used to select the functionally newlinesimilar and structurally complimentary template hits to model the protein sequences. Protein newlinemodels sampled through MODELLER are evaluated through different assessment scores viz. newlineMOLPDF, GA341, DOPE Score, Normalized DOPE Score, GDT-TS, GDT-HA and newlineTM_Score. TM_Score along with Normalized DOPE score (Z_Score) is lastly selected as the newlinebest set of model assessment measures and is employed to evaluate the model sampling for newlineselecting the accurate target model. The statistical ranking based template selection and newlinecombination algorithm, further integrated with TM_Score and Z_Score assessed iterative newlinesampling strategy, significantly improves the modelling accuracy of the targets. The newlinealgorithm predicts accurate models with an average GDT-TS, GDT-HA and TM_Score newlineimprovement of 3.531, 4.814 and 0.022 along with the individual relevant standard deviationPage vof 4.142,