FiCoS

FiCoS simulates large-scale biochemical networks using GPU-accelerated deterministic ODE integration to enable efficient analysis of mechanistic and rule-based models.

Key Features:

• Deterministic ODE simulation: Performs deterministic simulation of biochemical networks formulated as systems of ordinary differential equations (ODEs).
• Integration methods: Implements Dormand–Prince (DOPRI5) for non-stiff systems and Radau IIA (RADAU5) for stiff systems, both adaptive Runge–Kutta techniques of order 5.
• Adaptive step-size control: Uses dynamic adjustment of integration step-sizes during ODE resolution to control error and efficiency.
• GPU parallelization: Employs both fine-grained and coarse-grained parallelization on Graphics Processing Units (GPUs) to accelerate computations.
• Scalability: Targets models with hundreds or thousands of molecular species and reactions, as common in rule-based modeling.
• Performance: Can achieve computational speedups of up to 855× compared to other deterministic simulators.

Scientific Applications:

• Large-scale network simulation: Simulation and analysis of complex cellular processes represented by extensive biochemical networks.
• Model calibration: Supports extensive model calibration workflows requiring many deterministic simulations.
• Perturbation testing: Enables high-throughput perturbation testing of mechanistic and rule-based models with large species/reaction counts.
• Hypothesis exploration and experimental design: Facilitates exploration of hypotheses and design of targeted laboratory experiments based on mathematical models.

Methodology:

FiCoS uses DOPRI5 (Dormand–Prince) for non-stiff ODEs and RADAU5 (Radau IIA) for stiff ODEs—both adaptive order-5 Runge–Kutta methods with dynamic step-size control—and applies fine- and coarse-grained GPU parallelization for deterministic ODE simulation.