FiCoS: A fine-grained and coarse-grained GPU-powered deterministic simulator for biochemical networks

Area of Science:

• Computational Biology
• Systems Biology
• Biophysics

Background:

• Mathematical models are crucial for understanding cellular processes and generating testable hypotheses.
• Large-scale biochemical network models present computational challenges due to numerous species and reactions.
• Current simulation approaches struggle with the computational demands of complex models, limiting analysis.

Purpose of the Study:

• To develop a novel simulation approach that overcomes the computational limitations of current methods for biochemical network models.
• To introduce FiCoS, a "black-box" deterministic simulator designed for efficient parallelization on Graphics Processing Units (GPUs).
• To enhance the speed and effectiveness of analyzing large-scale, complex biological systems.

Main Methods:

• Implemented FiCoS, a novel deterministic simulator featuring fine-grained and coarse-grained parallelization on GPUs.
• Integrated two distinct numerical integration methods: Dormand-Prince for non-stiff systems and Radau IIA for stiff systems of Ordinary Differential Equations (ODEs).
• Evaluated FiCoS performance against existing deterministic simulators using models of varying sizes and computational loads.

Main Results:

• FiCoS demonstrated significant computational speedups, achieving up to 855× acceleration compared to other deterministic simulators.
• The simulator efficiently handled both non-stiff and stiff systems of coupled ODEs.
• Performance was validated across a range of model complexities and increasing computational demands.

Conclusions:

• FiCoS offers a powerful and efficient solution for simulating and analyzing large-scale biochemical networks.
• The GPU-based parallelization and dual integration methods enable dramatic acceleration of complex biological model computations.
• This approach promises to make the modeling of complex biological processes more effective and less computationally prohibitive.