Predicting overlapping protein complexes based on core-attachment and a local modularity structure.

Area of Science:

• Computational Biology
• Bioinformatics
• Systems Biology

Background:

• Detecting protein complexes (PCs) from protein-protein interaction networks (PPINs) is crucial in biological research.
• Existing graph clustering methods often overlook the core-attachment organization and local modularity of PCs.
• Limitations include poor seed selection, inadequate handling of dense subgraphs, and insufficient noise tolerance, leading to inaccurate predictions.

Purpose of the Study:

• To propose a novel clustering method, CALM, to address the limitations of existing approaches for detecting overlapping PCs.
• To incorporate core-attachment organization and local modularity structure for improved PC identification.
• To enhance the accuracy and robustness of PC detection in the presence of noise in weighted PPINs.

Main Methods:

• CALM identifies overlapping and seed nodes within weighted PPINs.
• It calculates a support function to recursively extend clusters from seed nodes based on local modularity.
• The method refines predictions through merging and removal procedures, ensuring robustness against noise.

Main Results:

• CALM demonstrates superior performance compared to classical methods in detecting overlapping PCs.
• It achieves higher accuracy and a better one-to-one mapping with reference complexes across multiple datasets.
• The method exhibits robustness against high rates of noise in PPINs.

Conclusions:

• CALM effectively detects protein complexes by integrating core-attachment and local modularity principles.
• The method offers improved accuracy and robustness over existing representative techniques.
• CALM has the potential to uncover novel overlapping PCs with diverse densities and high modularity.