How and when should interactome-derived clusters be used to predict functional modules and protein function?

Area of Science:

• Systems biology
• Bioinformatics
• Computational biology

Background:

• Protein-protein interaction (PPI) network clustering is widely used for predicting protein functions and modules.
• The effectiveness of these clustering approaches for functional analysis remains an open question.

Purpose of the Study:

• To develop a framework for assessing the performance of network clustering algorithms in predicting protein functions.
• To compare the efficacy of various clustering methods against a guilt-by-association approach for functional annotation.

Main Methods:

• Developed a general framework to evaluate the overlap between computationally derived clusters and Gene Ontology (GO) functional modules.
• Evaluated six diverse network clustering algorithms on the Saccharomyces cerevisiae interactome.
• Compared clustering-based function prediction with a guilt-by-association method.

Main Results:

• Clustering algorithm performance varied significantly on the same network and depended on network topology.
• A simple guilt-by-association approach surprisingly outperformed clustering-based methods for function prediction in S. cerevisiae.
• Performance analysis provided guidelines on when clustering approaches are most suitable for interactome analysis.

Conclusions:

• Results suggest a re-examination of the application of clustering to physical interactomes.
• Established guidelines for justifying and evaluating novel clustering approaches for biological network functional analysis.