Fast and flexible gpu accelerated binding free energy calculations within the amber molecular dynamics package

Area of Science:

• Computational chemistry
• Biophysics
• Drug discovery

Background:

• Alchemical free energy (AFE) calculations using molecular dynamics (MD) are vital for understanding biological processes and designing therapeutics.
• Current computational limitations, particularly with CPU-based simulations, hinder the routine application of AFE calculations in early-stage drug discovery due to long convergence times.
• The lengthy simulation times for AFE calculations on large systems (e.g., 50,000 atoms) restrict their use for analyzing numerous potential drug candidates.

Discussion:

• This study introduces a seamless Graphics Processing Unit (GPU) implementation of thermodynamic integration (TI) within the AMBER molecular dynamics package.
• The GPU-accelerated implementation significantly enhances computational efficiency, achieving speeds over 140 ns/day for large systems (44,907 atoms).
• The accuracy of the GPU implementation is validated to be equivalent to the existing CPU-based methods in AMBER.

Key Insights:

• Accelerated AFE calculations enable routine application in early-stage drug discovery.
• GPU implementation dramatically reduces simulation time for complex molecular systems.
• Accurate and rapid free energy calculations provide crucial insights into molecular recognition and druggability.

Outlook:

• The developed GPU implementation is integrated into the AMBER 18 package, making it accessible for broader research use.
• This advancement is expected to expedite the identification and optimization of novel therapeutic compounds.
• Facilitates deeper understanding of molecular interactions critical for treating various diseases.