A discard-and-restart MD algorithm for the sampling of protein intermediate states

Area of Science:

• Computational Biology
• Biophysics
• Drug Discovery

Background:

• Accurate sampling of protein intermediate states is crucial for understanding protein function and designing drugs.
• Traditional molecular dynamics (MD) simulations are often computationally expensive for exploring complex conformational landscapes.

Purpose of the Study:

• To introduce a novel discard-and-restart MD algorithm for efficient sampling of protein intermediate states.
• To enhance computational structure-based drug discovery by reducing simulation times.
• To enable the identification of druggable sites on dynamic protein targets.

Main Methods:

• Iterative short MD simulations with a collective variable loss function to guide trajectory progression.
• Discarding simulations that deviate from the target state and restarting with new initial velocities.
• AI-based analysis of transitory conformations to identify potential binding pockets.

Main Results:

• The algorithm achieved up to 2000× reduction in simulation time for protein folding pathways.
• Demonstrated efficacy in capturing folding pathways for small proteins, prion protein, and α-tubulin.
• Successfully identified potential binding pockets on dynamic protein states.

Conclusions:

• The discard-and-restart MD algorithm provides a computationally efficient method for exploring protein conformational landscapes.
• This approach significantly enhances the ability to discover ligands targeting dynamic protein states.
• The method holds promise for accelerating structure-based drug discovery and understanding protein dynamics.