ddcMD

ddcMD performs GPU-accelerated molecular dynamics simulations using the Martini coarse-grained force field for large-scale biomolecular systems.

Key Features:

• Martini coarse-grained force field: Implements the Martini coarse-grained force field for biomolecular simulations.
• GPU Acceleration: Fully GPU-accelerated execution achieves up to 278-fold speedup when comparing one GPU core to one CPU core.
• Heterogeneous programming model: Uses a heterogeneous GPU/CPU model that offloads intensive calculations to GPUs while using minimal CPU cores for I/O and orchestration.
• GPU-accelerated integration steps: Extends GPU acceleration to integration components including thermostat, barostat, and constraint solver functions.
• Performance benchmarking: Demonstrated performance includes 1.04 microseconds per day on a single NVIDIA V100 GPU and 6.19 microseconds per day across six GPUs on a Summit node for a 136,000-particle protein–lipid system.

Scientific Applications:

• Large-scale biomolecular simulations: Enables simulations of systems comprising tens to hundreds of thousands of particles to extend reachable timescales.
• Protein–lipid interactions: Suited for studying protein–lipid systems and their dynamic behavior in membranes.
• Membrane dynamics and macromolecular assemblies: Applicable to investigations of membrane dynamics and large macromolecular assemblies using coarse-grained models.

Methodology:

GROMACS input files are converted to a ddcMD-compatible format using the ddcMDconverter.