Dynamic Optimization with Particle Swarms (DOPS): a meta-heuristic for parameter estimation in biochemical models.

Area of Science:

• Biochemistry
• Computational Biology
• Systems Biology

Background:

• Mathematical modeling is crucial for designing biochemical networks.
• Parameter estimation in biochemical models is challenging due to expensive function evaluations, noisy data, and multiple local extrema.
• Existing methods struggle to efficiently optimize biochemical model parameters.

Purpose of the Study:

• To develop a novel hybrid meta-heuristic, Dynamic Optimization with Particle Swarms (DOPS), for efficient biochemical model parameter estimation.
• To address the challenges of parameter estimation in complex biochemical systems.
• To provide a robust and effective tool for analyzing biochemical networks.

Main Methods:

• DOPS combines multi-swarm particle swarm optimization with dynamically dimensioned search (DDS).
• It uses a multi-swarm approach to generate candidate solutions, with the best solution being refined by DDS.
• The method was tested on optimization test functions, biochemical benchmark problems, and a real-world model.

Main Results:

• DOPS demonstrated superior performance compared to differential evolution (DE), simulated annealing (SA), and DDS.
• The method achieved optimal solutions with a limited number of function evaluations.
• DOPS outperformed other meta-heuristics on test functions, benchmark problems, and the human coagulation cascade model.

Conclusions:

• DOPS is a promising meta-heuristic for biochemical model parameter estimation.
• The approach is effective in scenarios with limited function evaluations.
• DOPS offers an efficient solution for parameter estimation in complex biochemical systems.