PREDICTION OF ZINC-BINDING SITES IN PROTEINS BASED ON BAYESIAN METHOD

Zinc ranks the second metal ion in organism and zinc binding proteins play important roles in physiological processes. Due to the availability of protein sequence features, some well-known predictive tools had been developed, however, fewer researchers had focused on the integration of these tools. In view of this, an integrated predictor termed Bayes_Zinc based on Bayesian method for the prediction of zinc-binding sites has been presented. This method takes the prediction scores of three predictors taking sequences as the attributers of the classifier, with integrated classification model constructed based on Bayesian approach. The probabilities of positive and negative samples were calculated respectively and the sample was classified. When tested on a non-redundant dataset (Zhao_dataset), average MCC of our method reached 0.7256 on the whole internal, increased nearly by 4% -20% higher than the other three sequence-based methods. In order to further demonstrate the robustness and accuracy of our method, we also tested the two parameters recall and precision. The recall increased about 5% -11%, and the precision increased much more. The performance of Bayes_Zinc was better than the other three predictors by comprehensively analyzing. Our method can be used for the identification of the zinc-binding site on a large scale based on protein sequence, and it can also be used for the prediction of the metal protein function.