Automated multi-objective calibration of biological agent-based simulations

Area of Science:

• Computational Biology
• Systems Biology
• Immunological Modeling

Background:

• Agent-based simulation (ABS) is crucial for understanding complex biological systems, complementing laboratory experiments.
• Traditional calibration methods struggle with complex biological domains requiring multiple metrics for accurate characterization.
• Accurate simulation calibration is essential for reliable experimental controls and interpretation of results in simulation-based science.

Purpose of the Study:

• To develop and demonstrate a multi-objective optimization method for calibrating agent-based simulations against complex, multi-metric biological behaviors.
• To introduce Multi-Objective Calibration (MOC) for identifying optimal trade-offs in simulation parameter values across multiple objectives.
• To enhance the rigor and efficiency of the calibration process in biological simulations.

Main Methods:

• Developed a novel Multi-Objective Calibration (MOC) method based on multi-objective optimization principles.
• Applied MOC to calibrate a well-established immunological simulation model against both known and novel target behaviors.
• Implemented an overfitting detection mechanism within MOC to optimize computational efficiency.

Main Results:

• MOC successfully generated Pareto fronts, representing optimal trade-offs between simulation performance metrics for complex biological systems.
• Calibration using MOC demonstrated that simulation-derived conclusions exhibit broad robustness across different optimal parameter sets.
• The novel overfitting detection method effectively terminated MOC runs, saving computational resources without compromising solution quality.

Conclusions:

• MOC significantly improves the accuracy and speed of calibrating complex biological agent-based simulations.
• This approach enhances the reliability of simulations, leading to more informative biological predictions and identifying inadequate model representations.
• MOC provides a rigorous framework for simulation-based science, essential for advancing our understanding of intricate biological processes.