New ways to boost molecular dynamics simulations.

Area of Science:

• Computational Chemistry and Molecular Dynamics
• Biophysics and Structural Biology

Background:

• Dihydrofolate reductase (DHFR) is a common benchmark for molecular dynamics simulations.
• Efficient simulation of biomolecules requires optimized algorithms and computational hardware.

Purpose of the Study:

• To describe a set of algorithms for high-throughput molecular dynamics simulations.
• To analyze the performance of new algorithms on a single CPU, focusing on speed and reproducibility.
• To assess the impact of novel CPU instruction sets and simulation techniques.

Main Methods:

• Development and implementation of a mixed multiple time-step algorithm (5 fs timesteps).
• Tuned LINCS algorithm for bond angle constraints and fused pair list/force calculation.
• Pressure coupling using a 'densostat' and exploitation of AVX2 CPU instruction sets.

Main Results:

• Achieved simulation speeds of 160 nanoseconds/day for DHFR on a single Intel Core i7 CPU.
• Demonstrated reproducible trajectories with particle mesh Ewald (PME) and an 8.0 Å cutoff.
• Analyzed the performance impact of Intel's transactional memory, atomic instructions, and sloppy pair lists.

Conclusions:

• The described algorithms significantly enhance molecular dynamics simulation efficiency on CPUs.
• The algorithms are well-suited for GPU acceleration and can process diverse Protein Data Bank (PDB) files.
• An implementation is available within the YASARA molecular modeling and simulation program.