MOCASSIN-prot: a multi-objective clustering approach for protein similarity networks

Area of Science:

• Computational Biology
• Bioinformatics
• Evolutionary Biology

Background:

• Proteins contain conserved domains, and their evolution involves complex events like duplication, loss, and shuffling.
• Understanding protein evolution requires models integrating sequence divergence and domain content.
• Existing protein network construction methods may not fully capture evolutionary complexities.

Purpose of the Study:

• To adapt a game-theoretic approach for protein network construction.
• To extend this framework with multi-objective optimization and clustering refinement.
• To develop a novel method for clustering multi-domain proteins.

Main Methods:

• Adapted a game-theoretic approach into multi-objective optimization.
• Incorporated a clustering refinement procedure.
• Applied the MOCASSIN-prot method to cluster multi-domain proteins from ten genomes.

Main Results:

• MOCASSIN-prot utilizes both domain composition and quantitative sequence similarity.
• The method was compared to Markov clustering (TRIBE-MCL) and spectral clustering (SCPS).
• MOCASSIN-prot generated fewer false positives, more functionally coherent clusters, and better differentiated protein families.

Conclusions:

• MOCASSIN-prot offers improved accuracy in clustering multi-domain proteins.
• The method enhances the understanding of protein family evolution.
• MOCASSIN-prot provides a valuable tool for bioinformatics research.