MUNDO: protein function prediction embedded in a multispecies world.

Area of Science:

• Computational Biology
• Bioinformatics
• Systems Biology

Background:

• Protein function prediction is crucial for understanding biological systems.
• Network-based methods are powerful but can be limited by data availability within a single species.
• Conserved functional information across species offers a valuable resource for improving prediction accuracy.

Purpose of the Study:

• To introduce MUNDO, a novel cross-species co-embedding method for protein function prediction.
• To evaluate MUNDO's performance against existing methods using protein-protein interaction networks from different species.
• To demonstrate the benefits of leveraging conserved functional information across evolutionary scales.

Main Methods:

• Developed MUNDO, a method combining single-network and co-embedding techniques.
• Utilized protein-protein interaction (PPI) networks from human, mouse, and yeast species.
• Trained and tested MUNDO for predicting functional annotations in target species using data from model species.

Main Results:

• MUNDO achieved superior performance in predicting functional annotations for mouse and human networks when trained using cross-species data.
• The method also excelled in predicting functions in Baker's yeast when trained on Fission and Baker's yeast PPI networks.
• In a specific case (Fission yeast), using only the target species' network data performed best, highlighting context-specific nuances.

Conclusions:

• Cross-species information significantly enhances network-based protein function prediction.
• MUNDO offers a robust framework for integrating diverse species' data for improved functional annotation.
• The study underscores the utility of evolutionary conservation in computational biology and provides an open-source tool for researchers.