Complex Prediction in Large PPI Networks Using Expansion and Stripe of Core Cliques

Area of Science:

• Bioinformatics
• Computational Biology
• Systems Biology

Background:

• Protein-protein interaction (PPI) networks are crucial for understanding cellular organization and function.
• Network clustering is a key technique for analyzing PPI data and identifying functional modules.
• Accurate detection of protein complexes is essential for biological discovery.

Purpose of the Study:

• To develop a novel graph mining approach for detecting protein complexes within PPI networks.
• To improve the accuracy and efficiency of protein complex identification compared to existing methods.
• To leverage dense neighborhoods in interaction graphs for robust complex prediction.

Main Methods:

• The proposed method identifies size-3 cliques associated with each protein interaction (edge).
• These core cliques are expanded to form high-density subgraphs, representing potential protein complexes.
• Loosely connected proteins are removed, and redundancy is minimized using the Jaccard coefficient.

Main Results:

• The approach was evaluated on yeast and human PPI datasets.
• Predicted protein complexes showed significantly higher similarity to gold-standard datasets (CYC-2008, CORUM) than other methods.
• The technique demonstrates high efficiency in identifying biologically relevant protein complexes.

Conclusions:

• The novel graph mining approach effectively detects protein complexes by analyzing dense regions in PPI networks.
• This method offers improved accuracy and outperforms existing approaches in benchmark comparisons.
• The findings contribute to a better understanding of cellular organization and protein function through network analysis.