Direct coevolutionary couplings reflect biophysical residue interactions in proteins

Area of Science:

• Biophysics
• Computational Biology
• Structural Bioinformatics

Background:

• Homologous proteins exhibit sequence variability due to coevolutionary constraints.
• Direct-Coupling Analysis (DCA) infers protein structure by modeling these constraints.
• Current DCA methods map complex coupling matrices to scalar parameters, losing information.

Purpose of the Study:

• To investigate the quantitative biophysical information encoded within DCA coupling matrices.
• To explore the potential for enhanced protein structure inference from these matrices.

Main Methods:

• Spectral analysis of DCA-derived 21x21 coupling matrices from 70 protein families.
• Analysis of synthetic data from lattice-protein models for validation.
• Evaluation of sampling quality and regularization effects.

Main Results:

• DCA coupling matrices contain quantitative information on physico-chemical properties of amino acid interactions.
• Spectral analysis reveals biochemical features previously lost in scalar mapping.
• Results are robust across diverse protein families and validated with synthetic models.

Conclusions:

• The spectral properties of DCA coupling matrices offer a richer source of information for protein structure prediction.
• Understanding these biochemical features can improve DCA accuracy and interpretability.
• Further research into spectral analysis of coevolutionary data holds significant potential for structural biology.