Computational prediction of protein interaction networks through supervised classification techniques

Abstract

This paper implements integrative methods to predict Pairwise (PW) and Module-Based (MB) protein interactions in Saccharomyces cerevisiae. The predictive ability of combining diverse sets of relatively strong and weak predictive datasets is investigated. Different classification techniques: Naive Bayesian (NB), Multilayer Perceptron (MLP) and K-Nearest Neighbors (KNN) were evaluated. The assessment demonstrated that as the predictive power of single-source datasets became weaker, MLP and NB performed better than KNN. Generation of PPI maps for S. cerevisiae and beyond will be improved with new, higher-quality datasets with increased interactome coverage and the integration of classification methods.Keywords: Protein-Protein Interactions; PPIs; module-based interactions; machine and statistical learning; functional data; feature encoding; dataset integration; computational systems biology.Reference to this paper should be made as follows: Browne, F., Wang, H., Zheng, H. and Azuaje, F. (2008) 'Computational prediction of protein interaction networks through supervised classification techniques', Int.