Table 4.

Computational methods for prediction of protein-protein interaction.

Technique	Algorithms	Strengths	Weaknesses	Organism	Reference
Phylogenetic	Cluster analysis, maximum likelihood, maximum parsimony, Bayesian inference	Provides information of selective environmental pressure	Difficult to estimate divergence of proteins	H. pylori, P. falciparum	Ratmann et al., 2007
Machine learning	Random forest, decision tree, k-nearest neighbors, bayesian, Neural networks, support vector machine	Simple to understand, accurate	Dependent of parameter settings and features, black-box predictor, large data set for training	Vibrio cholerae, P. aeruginosa	Nanni et al., 2012; Ehrenberger et al., 2015
Data mining	Named entity recognition, ID3, Computational of natural language processing, C4.5	Fast and process large volumes of information, good to focused list	It is sensitive to noise, require manually curation	H. pylori, Campylobacter jejuni	Bock and Gough, 2003
Topological	Power-law degree distribution, clustering coefficient	Common topological characteristics among species (small-world), comparison with random networks	False positives proportional to the size of the network, configuration of protein modules may vary	E. coli	Butland et al., 2005; Wuchty, 2006; Sharan et al., 2007
Structure	Shape complementarity, rigid-body docking, heuristic potential	Accurate, good availability of data for primary and secondary structure	Slow development for high throughput methodologies	E. coli, S. typhimurium and T. maritima	Matsuzaki et al., 2014