Data Mining: Practical Machine Learning Tools and Techniques, Second Edition (Morgan Kaufmann Series in Data Management Systems)
CReF: a central-residue-fragment-based method for predicting approximate 3-D polypeptides structures
Proceedings of the 2008 ACM symposium on Applied computing
A Hybrid Method for the Protein Structure Prediction Problem
BSB '08 Proceedings of the 3rd Brazilian symposium on Bioinformatics: Advances in Bioinformatics and Computational Biology
Expert Systems with Applications: An International Journal
Mining the Protein Data Bank with CReF to predict approximate 3-D structures of polypeptides
International Journal of Data Mining and Bioinformatics
Expert Systems with Applications: An International Journal
Towards designing modular recurrent neural networks in learning protein secondary structures
Expert Systems with Applications: An International Journal
Hi-index | 12.05 |
Protein structure prediction (PSP) is a long standing problem in structural biology and bioinformatics. Within the PSP problem loop refinement is a major bottleneck. In this article we report the latest version of the CReF expert predictor system for the PSP problem with emphasis on loop refinement of the approximate 3-D structure 1ZDD_P of the Z34C mini protein predicted by CReF. We designed a loop refinement protocol based on seven molecular dynamics (MD) simulations runs at different temperatures. We found that, by letting the loop residues move freely during dynamics at 325K and restraining the internal coordinates of the correctly predicted helical structures, while allowing them to move relative to each other, the refinement protocol was very effective in predicting an accurate loop conformation in the first 100ps of a 1000ps MD simulation. The quality of the predictions was confirmed by the RMSD between refined and experimental structures which varied from 0.6 to 1.3A. In addition, stereochemical analyses showed that 100% of all residues of the refined 1ZDD_P, including those in the loop, populates the most favorable core regions of the Ramachandran plot. Our study suggests that the proposed protocol may be suitable to refine more complex mini proteins with different classes and architectures.