NeBcon: protein contact map prediction using neural network training coupled with naïve Bayes classifiers

• DOI: 10.1093/bioinformatics/btx164
• Publication Date: 2017-03-22T00:55:07Z
• Authors: Baoji He; S M Mortuza; Yanting Wang; Hong-Bin Shen; Yang Zhang
• Keywords: CASP

Recent CASP experiments have witnessed exciting progress on folding large-size non-humongous proteins with the assistance of co-evolution based contact predictions. The success is however anecdotal due to requirement prediction methods for high volume sequence homologs that are not available most protein targets. Development efficient can generate balanced and reliable maps different type targets essential enhance rate ab initio structure prediction.We developed a new pipeline, NeBcon, which uses naïve Bayes classifier (NBC) theorem combine eight state art built from machine learning approaches. posterior probabilities NBC model then trained intrinsic structural features through neural network final map prediction. NeBcon was tested 98 non-redundant proteins, improves accuracy best meta-server predictor by 22%; magnitude improvement increases 45% hard lack in databases. Detailed data analysis showed major contribution optimized combination complementary information both training also helps improve coupling probability features, were found particularly important do sufficient number homologous sequences derive profiles.On-line server standalone package program at http://zhanglab.ccmb.med.umich.edu/NeBcon/ [e-mail] Bioinformatics online.