Boosting Free-Energy Perturbation Calculations with GPU-Accelerated NAMD

Area of Science:

• Computational chemistry
• Biophysics
• Molecular modeling

Background:

• Molecular dynamics (MD) simulations are crucial for understanding biological systems.
• Free-energy calculations, particularly free-energy perturbation (FEP), are vital for determining molecular properties.
• Accelerating these computationally intensive methods is essential for larger and more complex systems.

Purpose of the Study:

• To develop and implement a GPU-accelerated version of the free-energy perturbation (FEP) method within the NAMD simulation package.
• To evaluate the performance and accuracy of the new GPU-based FEP implementation compared to existing CPU-based methods.
• To provide a more efficient and cost-effective computational tool for free-energy calculations in molecular simulations.

Main Methods:

• Implementation of FEP calculations on graphics processing units (GPUs) within the NAMD molecular dynamics engine.
• Benchmarking the GPU implementation against the traditional CPU implementation using standard molecular systems.
• Performance analysis focusing on speedup factors and accuracy preservation.

Main Results:

• The new GPU-based FEP implementation in NAMD achieves approximately four times the speed of the CPU version with no loss in accuracy.
• An optimized single-GPU node implementation demonstrates a speedup of up to nearly 30 times compared to the CPU implementation.
• These advancements significantly reduce the computational cost of free-energy calculations.

Conclusions:

• The developed GPU-accelerated FEP method in NAMD offers a substantial performance improvement for molecular dynamics simulations.
• This provides a more accessible and cost-effective approach for researchers to perform accurate free-energy calculations, especially for large biological molecules.
• The enhanced FEP functionality is available in the NAMD 3.0 release, empowering the scientific community.