cupSODA

A CUDA-powered, coarse-grain deterministic simulator for mass-action kinetics models in systems biology.

cupSODA is a black-box deterministic simulator of biological systems that exploits the remarkable memory bandwidth and computational capability of modern GPUs.

Investigating the emergent dynamics of complex biological systems often requires executing a massive number of simulations under varying conditions (e.g., parameter sweeps or sensitivity analysis). cupSODA was designed to efficiently execute these large batches of simulations in parallel. It works by automatically deriving the system of Ordinary Differential Equations (ODEs) from a reaction-based mechanistic model defined by mass-action kinetics, and then solving them using the LSODA numerical integration algorithm.

GitHub stars License

CUDA / C++ Systems Biology

Technical Architecture

cupSODA is built directly in C++ and CUDA (83.5% CUDA codebase) to run efficiently on NVIDIA architectures. It is meant to be highly portable and can be compiled across supported architectures (GNU/Linux, Microsoft Windows, macOS) targeting specific GPU compute capabilities.

The software is executed directly from the command line, allowing researchers to specify the model input directory, thread/block distribution, output destinations, and advanced memory configurations to squeeze out maximum parallel performance from the hardware.


Research Highlights

Massively Parallel

Distributes independent simulation instances across thousands of GPU cores for unparalleled throughput.

Automated ODEs

Automatically translates high-level biochemical mass-action reactions into calculable ODEs.

LSODA Integration

Utilizes the robust LSODA algorithm to ensure high numerical stability for stiff and non-stiff systems.

Core Publication

2014

  1. GPU-accelerated simulations of mass-action kinetics models with cupSODA
    Marco S Nobile, Paolo Cazzaniga, Daniela Besozzi, and 1 more author
    The Journal of Supercomputing, 2014