From principal component to direct coupling analysis of coevolution in proteins: low-eigenvalue modes are needed for...

Area of Science:

• Computational Biology
• Protein Bioinformatics
• Statistical Mechanics

Background:

• Covariation in protein sequences offers insights into structure and function.
• Principal Component Analysis (PCA) and Direct Coupling Analysis (DCA) are common methods for analyzing residue covariation.

Purpose of the Study:

• To introduce the Hopfield-Potts model, bridging PCA and DCA for protein sequence analysis.
• To demonstrate the model's efficacy in predicting residue-residue contacts and structural information.

Main Methods:

• Developed the Hopfield-Potts model inspired by statistical physics of disordered systems.
• Utilized eigenmodes and eigenvalues of the residue-residue correlation matrix.
• Applied dimensional reduction to avoid overfitting and analyze smaller datasets.

Main Results:

• The Hopfield-Potts model accurately predicts residue-residue contacts.
• Fewer parameters are required compared to Direct Coupling Analysis (DCA).
• Low-eigenvalue modes, often ignored by PCA, reveal localized patterns crucial for 3D structural information.

Conclusions:

• The Hopfield-Potts model provides an effective and parameter-efficient approach for protein sequence analysis.
• It successfully integrates residue contact prediction with structural information recovery.
• The model highlights the importance of low-eigenvalue modes in understanding protein structure.