AWE-WQ: fast-forwarding molecular dynamics using the accelerated weighted ensemble.

Area of Science:

• Computational Chemistry
• Biophysics
• Software Engineering

Background:

• Traditional molecular dynamics (MD) methods struggle to compute reaction rates due to high energy barriers trapping systems in metastable states.
• Weighted Ensemble (WE) methods offer efficient conformational sampling and unbiased rate calculations but lack scalable implementations for complex biomolecular systems.

Purpose of the Study:

• To introduce AWE-WQ, a scalable implementation of the Weighted Ensemble (WE) algorithm for molecular dynamics simulations.
• To enable efficient and unbiased calculation of reaction rates in large biomolecular systems.

Main Methods:

• Developed AWE-WQ, a GPLv2 implementation of a WE algorithm utilizing the master/worker distributed computing WorkQueue (WQ) framework.
• AWE-WQ supports thousands of nodes, dynamic resource allocation, heterogeneous computing (CPU/GPU), and integration with arbitrary MD codes like GROMACS.
• Ensured all statistical calculations remain unbiased across distributed and heterogeneous computing environments.

Main Results:

• AWE-WQ demonstrated scalability to thousands of nodes with a peak aggregate performance of 1000 ns/h.
• Simulated a 34-residue protein for 1.5 ms, achieving folding and unfolded rates comparable in accuracy to a 200 μs GPU simulation.
• Successfully applied AWE-WQ to a complex biomolecular system, validating its performance and accuracy.

Conclusions:

• AWE-WQ provides a highly scalable and flexible platform for performing unbiased rate calculations in molecular dynamics.
• The implementation overcomes previous limitations in WE method scalability, making it applicable to significant biomolecular research.
• AWE-WQ facilitates accurate computation of reaction rates, advancing the study of molecular dynamics and protein folding.