Large-scale asynchronous and distributed multidimensional replica exchange molecular simulations and efficiency...

Area of Science:

• Computational chemistry and biophysics.
• Molecular dynamics simulations.
• High-performance computing.

Background:

• Replica exchange molecular dynamics (REMD) is crucial for exploring complex molecular systems.
• Traditional REMD simulations face challenges in large-scale, distributed, and heterogeneous computing environments.
• Efficient simulation methods are needed to overcome computational bottlenecks.

Purpose of the Study:

• To introduce and validate asynchronous REMD (ASyncRE) methods for large-scale simulations.
• To demonstrate the applicability of ASyncRE on diverse computing platforms, including grid and high-performance clusters.
• To improve the efficiency and accuracy of molecular dynamics simulations for binding free energy estimation.

Main Methods:

• Implementation of ASyncRE on NSF XSEDE clusters, BOINC networks, and World Community Grid.
• Extensive simulations of the beta-cyclodextrin-heptanoate host-guest system using 1D and 2D ASyncRE.
• Application of the binding energy distribution analysis method for absolute binding free energy calculations.

Main Results:

• ASyncRE successfully performed over 60 ms of aggregate simulations on distributed and clustered environments.
• The method allowed for accurate estimation of absolute binding free energies for the host-guest system.
• Optimizing molecular dynamics (MD) periods and exchange attempts per cycle achieved efficiency comparable to synchronous REMD.

Conclusions:

• ASyncRE provides a viable and efficient approach for large-scale REMD simulations across heterogeneous computing resources.
• The developed methods facilitate accurate free energy calculations, crucial for drug discovery and molecular design.
• Optimized ASyncRE parameters can mitigate overheads in distributed computing, enhancing simulation throughput.