Folding Proteins at 500 ns/hour with Work Queue.

Area of Science:

• Computational chemistry
• Molecular dynamics simulations
• Distributed computing

Background:

• Traditional molecular modeling methods face significant computational costs and poor scalability due to long simulation trajectories.
• Existing techniques often require extensive communication between computer nodes, limiting parallel processing efficiency.

Purpose of the Study:

• To introduce a new class of molecular modeling methods that rely on numerous short calculations with minimal inter-node communication.
• To evaluate the efficiency and scalability of Accelerated Weighted Ensemble Dynamics (AWE) for molecular simulations.

Main Methods:

• Implementation of Accelerated Weighted Ensemble Dynamics (AWE) using the Work Queue framework for task management.
• Application of AWE to an all-atom protein model (Fip35 WW domain).
• Utilization of heterogeneous computing resources, including cloud platforms, clusters, and grids across various architectures (CPU/GPU, 32/64bit).

Main Results:

• Demonstrated excellent scalability by leveraging diverse and dynamic computing environments.
• Achieved an aggregate sampling rate exceeding 500 ns/hour, a substantial improvement over typical single-process rates of 0.1 ns/hour.
• Showcased efficient distributed computing for molecular dynamics simulations.

Conclusions:

• Accelerated Weighted Ensemble Dynamics (AWE) offers a highly scalable and efficient approach to molecular modeling.
• The proposed method effectively utilizes heterogeneous and dynamic computing resources, significantly accelerating simulation speeds.
• This advancement has the potential to reduce computational barriers in molecular simulations.