cuda-sim

cuda-sim accelerates biochemical network simulations by translating ODE, SDE, and MJP algorithms to NVIDIA CUDA GPUs for high-throughput simulation in systems and synthetic biology.

Key Features:

• Parallelization: Optimized for NVIDIA CUDA GPUs to execute large numbers of independent simulations concurrently, achieving up to a 360-fold reduction in runtime versus single-CPU implementations.
• Algorithmic Support: Provides parallelized implementations of LSODA for ODE integration, Euler-Maruyama for SDE simulation, and the Gillespie algorithm for Markov jump process simulation.
• Model Specification: Accepts biochemical network models specified in Systems Biology Markup Language (SBML) or provided directly as CUDA code.
• Implementation: Implemented as a Python package that leverages CUDA for GPU execution of simulation kernels.

Scientific Applications:

• Systems biology: Enables large-scale simulation-based analysis of biochemical network dynamics.
• Synthetic biology: Supports design and testing of synthetic circuits through extensive simulation experiments.
• Parameter exploration and statistical analysis: Facilitates extensive parameter-space exploration and calculation of simulation-derived statistics.

Methodology:

cuda-sim translates LSODA, Euler-Maruyama, and Gillespie implementations into highly parallelized CUDA kernels that execute many independent simulation trajectories concurrently on NVIDIA GPUs, leveraging the GPU architecture for massive parallel execution of independent calculations.