Calibration & Validation

Calibrating and validating an agent-based model (ABM) before making predictions is essential for ensuring the model's biological realism and establishing predictive accuracy, here is why:

Calibration: This process involves tuning model parameters (e.g., cell proliferation rates, diffusion coefficients, etc.) to match observed biological or experimental data. Proper calibration ensures that the model's behaviour mimics the real-world system it aims to represent. Without calibration, the ABM might produce unrealistic or biologically irrelevant dynamics, limiting its applicability to real-life scenarios.
Validation: This step involves comparing the model's outputs to independent experimental or clinical data not used during calibration. Validation demonstrates that the model can reproduce observed phenomena under a range of conditions, giving confidence that its predictions for untested scenarios will be accurate.

Why should you calibrate and validate you ABM?

Calibration without validation risks overfitting, where the model reproduces specific datasets perfectly but fails to generalize to new situations. Validation ensures that the model captures fundamental system dynamics, rather than just matching a particular dataset.
Predictions from an uncalibrated or unvalidated ABM can lack scientific credibility. Calibration and validation provide a foundation for confidence in the model's results, ensuring that predictions are backed by rigorous testing.
The process of calibration and validation helps pinpoint which parameters most significantly influence the system's behaviour. This understanding is vital for designing meaningful simulations and identifying key drivers of biological phenomena.
In applications like liver regeneration or cancer recurrence, ABMs are used to predict outcomes under various conditions (e.g., different surgical approaches or drug regimens). Validating the model ensures that these predictions are reliable enough to inform experimental designs or clinical decisions.
Calibration and validation uncover limitations in the model's design or assumptions, helping to refine it further. A well-calibrated and validated ABM is more robust and versatile, capable of simulating a broader range of scenarios with confidence.

My Agent-Based Model

Working closely with MD Dr Camilo Julio Llamoza-Torres, we used the liver regeneration model to calibrate and validate the ABM.
The liver is an ideal scenario to test and validate an agent-based model (ABM) for several reasons:

Complex Regenerative Dynamics: The liver exhibits a unique capacity for regeneration after injury or partial removal (e.g., partial hepatectomy). This regenerative process involves intricate interactions between different cell types, growth factors, and signalling pathways, making it a perfect testbed for ABMs designed to simulate multicellular systems. On the other hand, different hepatectomies produce different responses. For instance, after a 30% partial hepatectomy, the liver will regenerate by cellular hypertrophy, whereas after a 70% partial hepatectomy, the liver will regenerate by cellular hypertrophy and proliferation.
Heterogeneous Cellular Microenvironment: The liver comprises various cell types, including hepatocytes, stellate cells, Kupffer cells, and endothelial cells, each playing distinct roles in regeneration. This heterogeneity allows the ABM to demonstrate its ability to model diverse cell types and their interactions within a shared environment.
Spatial Organization and Tissue Architecture: The liver's spatial organization, including lobular structures and zonation, provides an excellent framework to test spatial dynamics and cellular behaviours in 3D tissue simulations. The model's ability to capture off-lattice dynamics is particularly relevant here.
Role of Diffusion: The liver relies heavily on the diffusion of nutrients, oxygen, and signalling molecules, which are crucial for maintaining tissue homeostasis and driving regeneration. By incorporating diffusion processes, the ABM can validate its capability to handle molecular transport in a physiologically relevant context.
Physiological and Clinical Relevance: Liver regeneration has been extensively studied, providing a wealth of experimental and clinical data to compare with and validate the ABM's predictions. This data-driven validation is essential for establishing the model's accuracy and utility in simulating biological systems.

By applying the ABM to liver regeneration, and comparing its results against experimental data, we assessed its ability to capture the complexity, heterogeneity, and dynamic processes of tissue systems, demonstrating its robustness and potential for broader applications.