Accuracy improvement in protein complex prediction from protein interaction networks by refining cluster overlaps

Area of Science:

• Computational Biology
• Bioinformatics
• Systems Biology

Background:

• Genome-wide protein-protein interaction (PPI) data analysis is advancing with computational techniques.
• Graph-clustering algorithms are used to predict protein complexes from PPI networks.
• Density-based clustering generates overlapping clusters, suitable for multi-complex protein membership, but accuracy is limited by overlap patterns.

Purpose of the Study:

• To develop a systematic approach for refining overlapping clusters from PPI networks.
• To introduce novel metrics for assessing cluster overlaps: overlap coverage and overlapping consistency.
• To improve the accuracy of protein complex prediction using these refined clusters.

Main Methods:

• A novel overlap refinement algorithm was designed, taking existing density-based clusters as input.
• The algorithm utilizes newly defined metrics (overlap coverage, overlapping consistency) to parameterize refinement.
• Protein complex prediction accuracy was evaluated using the f-measure against known complexes.

Main Results:

• The proposed method refines overlapping clusters identified from protein interaction networks.
• Experimental results on yeast PPI data (BioGRID, DIP) show significant accuracy increases in protein complex prediction.
• The refined clusters demonstrate improved performance compared to existing methods.

Conclusions:

• The cluster overlap refinement approach is effective for protein complex detection.
• Analyzing cluster overlaps in PPI networks is vital for understanding protein functions and system topology.
• This method enhances the biological insights derived from PPI network analysis.