PPIs Meta: A Meta-predictor of Protein-Protein Interaction Sites with Weighted Voting Strategy

Author(s): Xiaowei Zhao, Lingling Bao, Xiaosa Zhao, Minghao Yin*.

Journal Name: Current Proteomics

Volume 14 , Issue 3 , 2017

Become EABM
Become Reviewer

Graphical Abstract:


Background: Meta-prediction seeks to combine the strengths of multiple predicting programs with the hope of improving the predicting performance surpassing that of all existing predictors. As increasing numbers of predicting programs have been developed in this field, there is an urgent need for effective meta-prediction strategies.

Objective: In this paper, we apply meta-prediction for the identification of protein-protein interaction (PPI) sites which is still a challenge task in computational biology.

Method: In this paper, six PPI sites predictors, CRF, SSWRF, DC-RF-RUS-PF, LORIS, SPPIDER and SPRINGS, are integrated to construct meta-predictors using several methods, including unweighted voting, unreduced weighted voting, reduced unweighted voting and weighted voting strategies.

Results: PPIsMeta, the meta-predictor produced by using weighted voting strategy with parameters selected by restricted grid search, performs the best for predicting the PPI sites. Its accuracy and Matthew’s Correlation Coefficient reach up to 70.6% and 0.26, respectively. Compared to the best individual element predictor (SSWRF), the predictor proposed in this study achieves a significant improvement in Matthew’s Correlation Coefficient of 0.036, and an improvement in accuracy of 5.8%.

Conclusion: The experimental results demonstrate that PPIsMeta is a powerful tool for predicting PPI sites. Compared with the existing methods, we find that our tool shows greater robustness in accuracy and Matthew’s Correlation Coefficient. PPIsMeta is available to the public at

Keywords: Protein-protein interaction sites, element predictor, meta-prediction, ensemble, voting strategy, hybrid.

Rights & PermissionsPrintExport Cite as

Article Details

Year: 2017
Page: [186 - 193]
Pages: 8
DOI: 10.2174/1570164614666170306164127
Price: $25

Article Metrics

PDF: 19