Autoencoder-Based Detection of Dynamic Allostery Triggered by Ligand Binding Based on Molecular Dynamics

Area of Science:

• Biophysics
• Computational Biology
• Structural Biology

Background:

• Dynamic allostery is crucial for protein function, involving information transmission via subtle dynamic changes rather than large conformational shifts.
• Understanding protein regulation mechanisms remains challenging despite extensive experimental and theoretical research.

Purpose of the Study:

• To propose and validate a novel autoencoder-based method for detecting dynamic allostery in proteins.
• To investigate the impact of ligand binding on protein dynamics and correlated motions.

Main Methods:

• Utilized molecular dynamics simulations to obtain protein structure fluctuations in ligand-bound and -unbound states.
• Employed an autoencoder model to analyze distance matrices derived from simulation data.
• Compared the autoencoder's findings with Principal Component Analysis (PCA) and Dynamic Cross-Correlation Matrix (DCCM) methods.

Main Results:

• The autoencoder method successfully detected dynamic allostery in the PDZ2 domain.
• Identified reorganization of correlative fluctuation motions among residue pairs upon ligand binding.
• Revealed distinct correlated motions compared to traditional PCA and DCCM analyses.

Conclusions:

• The proposed autoencoder method offers a new perspective on understanding protein allosteric regulation by focusing on dynamic changes.
• This approach can be valuable for studying signal transduction, mutagenesis, and diseases linked to protein dysfunction.