Scaling molecular dynamics beyond 100,000 processor cores for large‐scale biophysical simulations

Abstract

The growing interest in the complexity of biological interactions is continuously driving the need to increase system size in biophysical simulations, requiring not only powerful and advanced hardware but adaptable software that can accommodate a large number of atoms interacting through complex forcefields. To address this, we developed and implemented strategies in the GENESIS molecular dynamics package designed for large numbers of processors. Long-range electrostatic interactions were parallelized by minimizing the number of processes involved in communication. A novel algorithm was implemented for nonbonded interactions to increase single instruction multiple data (SIMD) performance, reducing memory usage for ultra large systems.Memory usage for neighbor searches in real-space nonbonded interactions was reduced by approximately 80%, leading to significant speedup. Using experimental data describing physical 3D chromatin interactions, we constructed the first atomistic model of an entire gene locus (GATA4). Taken together, these developments enabled the first billion-atom simulation of an intact biomolecular complex, achieving scaling to 65,000 processes (130,000 processor cores) with

New parallel computing algorithm of molecular dynamics for extremely huge scale biological systems

“…Table 1 on Oakforest-PACS. 3 The performance of GENESIS on Fugaku is also better than efforts using other MD software for simulating a 1B atoms system. In our understanding, the previous best performance for 1B atoms system so far is 5 ns/day using NAMD on 16,384 nodes of Oak Ridge Titan GPUs.…”

Antimicrobial Metal Nanomaterials: From Passive to Stimuli‐Activated Applications

“…In classical atomistic MD simulations, molecules are typically represented by atomic beads with a fixed charge that are connected by bonds, angles, and dihedrals, while intermolecular interactions are described by electrostatic and van der Waals terms. While this simplistic treatment allows for simulations of up to billions of atoms [ 143 ] and up to millisecond timescales, [ 144 ] because bonds and atomic partial charges are fixed, chemical reactions, optical or electronic properties, and the effects of polarization cannot be thoroughly examined. Nevertheless, MD methods have proven useful in studying the bacterial cell wall, [ 145 ] elucidating the process of fungal biofouling [ 146 ] and providing insight into antifouling materials, [ 147 ] and by providing design principles for photoluminescent nanoparticles, [ 148 ] among others.…”

How Do Molecular Dynamics Data Complement Static Structural Data of GPCRs

“…In recent years, there has been a dramatic increase in achieved timescales. This tendency is expected to continue, thanks to the advances in algorithms [ 8 , 9 , 157 , 158 ], software [ 159 , 160 , 161 , 162 ], and hardware [ 163 , 164 ] that we are experiencing. In fact, it is expected that all-atom, classical MD simulations will be able to reach the second timescale within the next five years [ 165 , 166 , 167 ].…”