A unified statistical model of protein multiple sequence alignment integrating direct coupling and insertions

Area of Science:

• Computational Biology
• Bioinformatics
• Statistical Mechanics

Background:

• Multiple sequence alignments (MSAs) reveal residue conservation patterns, including short- and long-range correlations.
• Existing models often struggle to incorporate both correlation types and insertions effectively.

Purpose of the Study:

• To develop a statistical model for MSAs that captures short-range and long-range residue correlations, as well as insertions.
• To investigate the impact of long-range interactions on protein conservation patterns and structural stability.

Main Methods:

• Derived a lattice gas model for MSAs based on the principle of maximum entropy.
• Employed the transfer matrix method with a mean-field approximation to compute the partition function.
• Determined model parameters using self-consistent conditions and Gaussian approximation.

Main Results:

• The model successfully accounts for all possible alignments and sequences.
• Analysis of globin and V-set domains showed that long-range interactions increase conservation pattern specificity to protein structure.
• Long-range interactions were found to increasingly stabilize well-conserved residues.

Conclusions:

• The developed lattice gas model provides a robust statistical framework for analyzing MSAs.
• Long-range interactions play a crucial role in refining residue conservation patterns and enhancing protein structural integrity.