STOCHSIMGPU: parallel stochastic simulation for the Systems Biology Toolbox 2 for MATLAB

Area of Science:

• Computational biology
• Systems biology
• Bioinformatics

Background:

• Stochasticity is crucial in biological systems, necessitating efficient simulation software.
• Existing computational methods for stochastic simulations are often slow.
• STOCHSIMGPU addresses the need for faster, biologically realistic stochastic simulations.

Purpose of the Study:

• To introduce STOCHSIMGPU, a novel software tool for accelerating stochastic simulations.
• To demonstrate the efficiency gains of using graphics processing units (GPUs) for these simulations.
• To provide an open-source, integrated solution for MATLAB users.

Main Methods:

• Developed STOCHSIMGPU, a software tool leveraging GPUs for parallel processing.
• Implemented the Gillespie stochastic simulation algorithm (SSA), logarithmic direct method (LDM), and next reaction method (NRM) on GPUs.
• Integrated STOCHSIMGPU with MATLAB and Systems Biology Toolbox 2 (SBTOOLBOX2).

Main Results:

• Achieved an approximately 85-fold speedup compared to sequential central processing unit (CPU) implementations.
• Demonstrated significant efficiency gains in parallel stochastic simulations.
• Ensured seamless integration by acting as a direct replacement for existing SBTOOLBOX2 stochastic simulation software.

Conclusions:

• STOCHSIMGPU offers substantial performance improvements for stochastic simulations.
• The software requires no model modifications, facilitating easy adoption.
• Open-source availability promotes wider use and development in computational biology.