PPalign

PPalign aligns Potts models of proteins to compute optimal pairwise alignments that incorporate direct residue–residue couplings for improved detection of remote homology and structural and functional relationships.

Key Features:

• Potts Model Representation: Represents proteins as Potts models encoding positional fields and direct residue–residue couplings that capture coevolutionary patterns.
• Integer Linear Programming (ILP) Formulation: Formulates alignment of Potts models as an integer linear programming problem to compute optimal pairwise alignments.
• Performance and Efficiency: Assessed on a non-redundant set of reference pairwise sequence alignments from the SISYPHUS benchmark focusing on low-identity sequences and achieves alignments with reasonable computation times.
• Comparative Advantage: Outperforms HHalign and a version of PPalign that omits couplings in comparative evaluations, yielding superior alignment scores.
• Research Implications: Demonstrates the potential of Potts-model-based alignments to improve sequence homology detection and supports unbiased investigations through guaranteed optimality.

Scientific Applications:

• Structural annotation: Integrates coevolutionary coupling information to aid structural annotation of proteins.
• Functional annotation: Supports functional annotation by aligning conserved coupling patterns indicative of functional similarity.
• Remote homology detection: Enhances detection of evolutionary relationships and remote homology among proteins with low sequence identity by leveraging pairwise couplings.

Methodology:

Represents proteins as Potts models encoding positional fields and direct couplings and aligns these models using an integer linear programming (ILP) formulation; performance was evaluated using non-redundant SISYPHUS reference pairwise alignments focusing on low-identity sequences.