DeepConPred2: An Improved Method for the Prediction of Protein Residue Contacts.

ABSTRACT: Information of residue-residue contacts is essential for understanding the mechanism of protein folding, and has been successfully applied as special topological restraints to simplify the conformational sampling in de novo protein structure prediction. Prediction of protein residue contacts has experienced amazingly rapid progresses recently, with prediction accuracy approaching impressively high levels in the past two years. In this work, we introduce a second version of our residue contact predictor, DeepConPred2, which exhibits substantially improved performance and sufficiently reduced running time after model re-optimization and feature updates. When testing on the CASP12 free modeling targets, our program reaches at least the same level of prediction accuracy as the best contact pre

SUBMITTER: Ding W 

PROVIDER: S-EPMC6247404 | biostudies-literature | 2018

REPOSITORIES: biostudies-literature