Evaluating Clustering Algorithms for Protein-Protein Interaction Networks
Author Information
Author(s): Brohée Sylvain, van Helden Jacques
Primary Institution: Université Libre de Bruxelles
Hypothesis
How do different clustering algorithms perform in extracting protein complexes from interaction networks?
Conclusion
The study concludes that the Markov Clustering (MCL) algorithm is the most robust method for extracting protein complexes from interaction networks.
Supporting Evidence
- MCL showed remarkable robustness to graph alterations.
- RNSC was more sensitive to edge deletion but less sensitive to suboptimal parameters.
- MCODE and SPC performed weaker under most conditions.
Takeaway
Scientists tested four different methods to group proteins that work together, and found that one method, called MCL, works best at finding these groups even when the data is messy.
Methodology
The study involved creating a test graph from annotated protein complexes and applying four clustering algorithms to evaluate their performance under various conditions.
Potential Biases
The study may be biased towards algorithms that are easier to use or configure.
Limitations
The evaluation was performed by users who may not be experts in the algorithms, potentially underestimating their capabilities.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website