Integrative computing method for the prediction of zinc-binding sites in proteins

A large number of metalloproteins contained in Protein Data Bank, taking metal ions as cofactors, have important biological functions. As the second most abundant bound trace metal elements in organism, zinc ion plays an important regulatory role in the biological growth and development, disease control, DNA synthesis. So, research on the area of zinc-binding protein sites has an important significance. Because of the availability of protein sequence information, a series of predictive tools based on sequence for zinc-binding sites in proteins have been developed. But presently, there is little research on the integration of these tools. Based on this, an integrative predictor termed meta-zincPrediction is presented in the paper to combine the predictive tools. The linear regression method is used in the integrated approach to combine the scores of different prediction tools, the parameters are adjusted and optimized until to the optimal. Tested on the non-redundant Zhao_dataset, AURPC value of meta-zincPrediction reached nearly 0.9, an increase of 2% -9% than other three predictors, and other performance indexes were also improved mostly. Moreover, the prediction ability of meta-zincPrediction was better than other three predictors, regardless of zinc-binding sites for all four types of residues, or zinc-binding sites for a single type of residues. Our method can not only be applied to large-scale identification of zinc-binding sites based on sequence information, but also be useful for the inference of protein function.