A multi-scale coevolutionary approach to predict interactions between protein domains

Area of Science:

• Computational Biology
• Bioinformatics
• Evolutionary Genomics

Background:

• Protein interactions are fundamental to biological processes.
• Genomic data offers vast resources for studying protein variability and interactions.
• Existing methods for predicting protein interactions have limitations.

Purpose of the Study:

• To develop and validate a novel computational approach for predicting protein interactions.
• To leverage multi-scale coevolutionary signals for enhanced prediction accuracy.
• To identify previously unknown direct interactions between protein families.

Main Methods:

• Introduction of 'direct phyletic couplings' based on global statistical models of phylogenetic profiles.
• Integration of direct coupling analysis with inter-protein residue-residue coevolution.
• Comparison of the new method against traditional correlation-based approaches like phylogenetic profiling.

Main Results:

• Direct phyletic couplings significantly increase the accuracy of predicting related protein domains (80% vs. 30-50% positives).
• Multi-scale coevolutionary evidence supports direct, yet previously unknown, interactions between protein families.
• Identified negative phyletic couplings reveal alternative functional solutions and convergent evolution.

Conclusions:

• Global statistical modeling offers powerful tools for genome-wide coevolutionary analysis.
• The developed methods provide biologically sensible and experimentally testable predictions of protein interactions.
• This work expands the application of coevolutionary analysis beyond individual protein complexes.