Identifying coevolving partners from paralogous gene families

Area of Science:

• Computational Biology
• Bioinformatics
• Evolutionary Biology

Background:

• Detecting coevolution typically requires a priori one-to-one mapping of sequence partners.
• Distinct duplication and loss histories in gene families complicate this mapping process.
• Existing methods are limited when dealing with complex phylogenetic relationships.

Purpose of the Study:

• To develop an algorithm for identifying coevolving partners between two sequence families with distinct phylogenetic trees.
• To overcome the limitations of existing methods that require predefined one-to-one mappings.
• To extend the applicability of phylogeny-based coevolutionary models.

Main Methods:

• Developed an algorithm that maps gene trees to a reference species tree.
• Constructed a joint state of sequence composition and coevolving partner assignments for each species tree node.
• Applied dynamic programming to the joint states for optimal assignment identification.

Main Results:

• The algorithm achieves 60%-88% accuracy in identifying coevolving partners on simulated and real-world protein domain data.
• Demonstrated the algorithm's effectiveness on sequences with distinct phylogenetic histories.
• Successfully identified coevolving partners without requiring a priori one-to-one mapping.

Conclusions:

• The proposed algorithm effectively identifies coevolving partners in sequence families with divergent evolutionary paths.
• This method enhances the capability of phylogeny-based coevolutionary analyses.
• The algorithm has broad applications in predicting molecular interactions, including protein-protein, protein-DNA, and DNA-RNA interactions.