seekrflow: Towards an End-to-End Automated Simulation Pipeline with Machine-Learned Force Fields for Accelerated...

Area of Science:

• Computational chemistry
• Molecular dynamics
• Biophysics

Background:

• Accurate prediction of drug-target binding kinetics and thermodynamics is crucial for drug discovery.
• Traditional atomistic simulations struggle with computational cost for rare binding events.
• Existing enhanced sampling methods require manual intervention, limiting scalability and introducing artifacts.

Purpose of the Study:

• To introduce seekrflow, an automated multiscale milestoning simulation pipeline.
• To streamline the prediction of drug-target binding kinetics and thermodynamics.
• To minimize manual intervention and computational overhead in simulations.

Main Methods:

• seekrflow employs an automated multiscale milestoning approach.
• The pipeline integrates steps from input structure to kinetic and thermodynamic predictions.
• Validation was performed on receptor-ligand complexes like HSP90, tyrosine kinase, and trypsin inhibitors.

Main Results:

• seekrflow accurately predicts kinetic and thermodynamic parameters for multiple receptor-ligand complexes.
• Predicted parameters closely match experimental estimates.
• The pipeline demonstrates efficiency and reproducibility.

Conclusions:

• seekrflow establishes a new benchmark for automated, high-throughput, physics-based predictions.
• The method enhances accuracy and reduces computational cost in drug discovery.
• seekrflow minimizes manual intervention, improving the reliability of kinetic and thermodynamic estimates.