PPalign: optimal alignment of Potts models representing proteins with direct coupling information.

Area of Science:

• Computational Biology
• Bioinformatics
• Structural Biology

Background:

• Protein annotation relies on homology search methods like BLAST and profile Hidden Markov Models.
• Current methods do not account for coevolutionary information between amino acid residues.
• Advances in contact prediction enable new modeling approaches for protein sequences.

Purpose of the Study:

• To develop a novel method for aligning protein sequences that incorporates coevolutionary information.
• To represent proteins using Potts models, which capture direct couplings between positions.
• To computationally address the challenge of aligning these complex Potts models.

Main Methods:

• Formulated the Potts model alignment problem using Integer Linear Programming.
• Developed PPalign, a program implementing this formulation for optimal pairwise alignment.
• Assessed PPalign performance on low-sequence-identity alignments from the SISYPHUS benchmark.

Main Results:

• PPalign successfully aligns Potts models in tractable time, averaging [Formula: see text] per alignment.
• The method demonstrates improved alignment accuracy compared to HHalign and PPalign without couplings.
• PPalign achieved better mean [Formula: see text] scores, particularly for distantly related sequences.

Conclusions:

• Pairwise couplings from protein Potts models enhance the alignment of remotely related sequences.
• Further research is needed to optimize Potts model inference for homology search.
• PPalign offers a robust tool for investigating Potts models and improving protein sequence alignment.