Evolutionary analysis and interaction prediction for protein-protein interaction network in geometric space

Area of Science:

• Systems Biology
• Bioinformatics
• Computational Biology

Background:

• Protein-protein interactions (PPIs) are fundamental to cellular processes.
• Understanding PPI networks is crucial for systems biology.
• Leveraging network-level information, including evolutionary insights, can improve PPI prediction.

Purpose of the Study:

• To develop a novel method for PPI prediction by integrating evolutionary information into a geometric space.
• To enhance the accuracy of predicting pairwise PPIs.
• To provide a framework for selecting and evaluating evolutionary models for PPI networks.

Main Methods:

• Incorporation of evolutionary information into a geometric space for PPI prediction.
• Cross-validation using human and yeast PPI network data.
• Development and testing of a modified evolutionary model (DANEOsf) combining gene duplication/neofunctionalization and scale-free models.

Main Results:

• Achieved up to a 14.6% increase in PPI prediction accuracy (measured by ROC score) compared to a baseline without evolutionary information.
• The DANEOsf evolutionary model demonstrated superior fitness and prediction efficacy on human and yeast PPI networks.
• The proposed method effectively selects evolutionary models that best represent underlying evolutionary mechanisms.

Conclusions:

• The novel method significantly improves PPI prediction accuracy by utilizing evolutionary information.
• The DANEOsf model offers a better fit for understanding the evolutionary mechanisms of human and yeast PPI networks.
• This work provides an effective approach for evaluating evolutionary models based on their PPI prediction performance.