GPU-accelerated Red Blood Cells Simulations with Transport Dissipative Particle Dynamics

Area of Science:

• Computational biology
• Biophysics
• Mesoscopic simulations

Background:

• Red blood cell (RBC) functionalities are influenced by chemical factors.
• Mesoscopic simulations offer a method to quantify these chemical influences.
• Existing simulation methods may lack the ability to capture simultaneous advection, diffusion, and reaction.

Purpose of the Study:

• To present a GPU-accelerated simulation package for mesoscopic RBC modeling.
• To adapt the transport Dissipative Particles Dynamics (tDPD) method for RBC simulations.
• To enable efficient investigation of chemical effects on RBCs.

Main Methods:

• Developed a GPU-accelerated simulation package using a tDPD adaptation.
• Implemented parallel processing for computational workloads on GPUs.
• Utilized multi-stream scheduling and non-blocking MPI for inter-node scalability.

Main Results:

• The simulation package accurately recovers RBC membrane properties (viscosity, elasticity, bending stiffness) and cross-membrane chemical transport.
• Achieved a speedup of 10.1 on a single GPU compared to 16 CPU cores.
• Demonstrated strong and weak scaling, with 91% weak scaling efficiency across 256 nodes.

Conclusions:

• The GPU-accelerated tDPD package provides accurate and efficient simulations of RBC behavior.
• Enables high-throughput investigation of chemical-driven RBC phenomena and disorders.
• Facilitates quicker understanding of RBC diseases through advanced computational modeling.