During the last years, high throughput experimental methods have been developed which generate large datasets of protein – protein interactions. However these datasets are prone to errors, reducing the quality of information they contain.

These datasets can be modeled as graphs, where vertices represent proteins and edges the pairwise protein – protein interactions, making it easy for automated clustering methods to be applied on them in order to detect protein complexes or other biological significant functional groups.  

A clustering tool, called GIBA (named by the first letters of its developers’ names), is presented. GIBA offers the users the ability to apply clustering algorithms in protein - protein interaction networks. GIBA implements a new two step methodology: In the first step, the protein interaction network is clustered by the Markov Clustering algorithm (MCL) or the restricted neighborhood search clustering algorithm (RNSC). In the second step, the clustering results are filtered to derive the final candidate protein complexes.

The efficiency of GIBA is demonstrated through experiments on 7 different yeast protein interaction datasets and comparisons with 4 other algorithms. For this purpose, the recorded yeast protein complexes of the MIPS database were used as benchmark and also 5 different metrics were used to evaluate the results: a) percentage of successful prediction, b) mean score of valid predicted complexes, c) sensitivity, d) positive predictive value and e) geometrical accuracy. Moreover, the parameters of the methods that constitute the filter have been checked on how they affect the final results.

GIBA is an effective and easy to use tool for the detection of protein complexes through clustering on protein - protein interaction networks. It was proved with extensive experiments that GIBA produces more quality approximations of protein complexes than other methods.

Description of the analysis made of GIBA filtering methods

Click here to view the result table.

Download the GIBA-tool source code

The datasets used in our experiments can be downloaded here.

The results of the tested algorithms can be downloaded here.