Prediction Of Protein Contact Map Based On Weighted Naive Bayes Classifier And Extreme Random Tree

The accurate prediction of residue-residue contacts provides crucial help to the ab initio protein folding and 3D structure modeling,because the accurately predicted contacts can enforce useful constraints to the structure assembly.Recent CASP experiments have witnessed the prosperities on-this-topic-and a number of promising protein contact map predictors have emerged in the past decades.Although much progress has been made,challenges(e.g.,low prediction accuracy for long-range contacts)remain.In this paper,we developed a new meta-based predictor,called TargetPCM,which can achieve high accuracy for protein contact map prediction(especially for medium-and long-range contacts).Firstly,TargetPCM tries to simplify and select three existing powerful contact map predictors,and combines the outputs of these predictors by using a weighted naive Bayes classifier(WNBC),among which the weight parameters are optimized with particle swarm optimization(PSO)algorithm,and generates the predictor-based features;then,TargetPCM tries to extract more effective sequence information to generate the sequence-based features,further combines them with the predictor-based features to form the discriminative features;finally,the extracted features are fed to the final prediction model,which is trained with extremely randomized trees(ERT),for performing contact map prediction.In order to verify the effectiveness of the method proposed in protein contact map prediction.Tested on two benchmark datasets,for short-,medium-and long-range contacts,the Top L/5 accuracy of our TargetPCM is better than other existing powerful contact map predictors.Moreover,further investigations on the performance at different coverage cutoffs and comparison of the true and predicted contact maps,detailed analysis on the experimental results shows that both the effective utilization of complementary information from base predictors and the powerful learning capability of ERT account for the performance improvements of the proposed TargetPCM over existing contact map predictors.