Using information theory to search for co-evolving residues in proteins

Area of Science:

• Bioinformatics
• Computational Biology
• Molecular Evolution

Background:

• Functionally critical protein residues are often conserved in multiple sequence alignments (MSAs).
• Compensatory mutations can mask functionally important residues, making them difficult to identify.
• Distinguishing co-evolving sites from non-conserved sites is a significant challenge in protein analysis.

Purpose of the Study:

• To develop and evaluate methods for identifying co-evolving protein residues.
• To assess the impact of sequence number, alphabet size, and mutation rate on co-evolving site detection.
• To determine the optimal normalization strategy for Mutual Information (MI) in detecting co-evolving sites.

Main Methods:

• Utilized Mutual Information (MI) to identify co-evolving positions in protein sequences.
• Employed in silico evolved MSAs to simulate various evolutionary conditions.
• Assessed the performance of different MI normalization techniques, including normalization by pair entropy.
• Analyzed real protein alignments to validate the findings.

Main Results:

• Normalization by pair entropy was found to be the optimal method for enhancing the detection of co-evolving positions.
• The study examined the influence of sample size, amino acid alphabet size, and mutation rate on background MI.
• Analysis of real protein alignments revealed that co-evolving residue pairs are frequently in contact.

Conclusions:

• Mutual Information normalized by pair entropy is a robust method for identifying co-evolving protein residues.
• This approach can help uncover functionally important residues that might otherwise be missed due to compensatory mutations.
• The findings facilitate a deeper understanding of protein structure-function relationships and evolutionary dynamics.