Predicting protein-protein interaction by the mirrortree method: possibilities and limitations

Area of Science:

• Bioinformatics
• Computational Biology
• Evolutionary Biology

Background:

• Molecular co-evolution analysis, a sequence-based method, is utilized for predicting protein-protein interactions.
• The mirrortree method, employing Pearson's correlation, quantifies evolutionary correlation between protein pairs.

Purpose of the Study:

• To evaluate the utility of the mirrortree method for predicting protein-protein interactions in eukaryotes.
• To assess how varying evolutionary distance and span metrics influence prediction accuracy.

Main Methods:

• Applied the mirrortree method to known interacting and non-interacting protein pairs.
• Analyzed protein conservation across diverse eukaryotic clades (mammals to fungi).
• Varied metrics for evolutionary distance and species evolutionary span.

Main Results:

• Co-evolutionary correlation scores showed significant predictive power for interacting versus non-interacting proteins when normalized for evolutionary span.
• Predictive accuracy was highest for proteins conserved across wide eukaryotic evolutionary ranges.
• Prediction accuracy was considerably weaker for narrower evolutionary spans.

Conclusions:

• The mirrortree method, particularly when normalized for evolutionary span, is a valuable tool for predicting protein-protein interactions in eukaryotes.
• The effectiveness of this method is contingent on the breadth of the evolutionary span analyzed, with wider spans yielding more robust predictions.