Topologically-based parameter inference for agent-based model selection from spatiotemporal cellular data.

Area of Science:

• Computational Biology
• Systems Biology
• Bioinformatics

Background:

• Spatiotemporal single-cell imaging generates complex data on cell dynamics and interactions.
• Extracting mechanistic insights from such data is challenging.
• Agent-based models (ABMs) simulate emergent population behavior from individual cell rules, while topological data analysis (TDA) describes spatial organization.

Purpose of the Study:

• To introduce TOPAZ, a computational pipeline for identifying plausible ABMs from spatiotemporal cellular data.
• To integrate TDA with approximate Bayesian computation (ABC) and Bayesian model selection for robust model inference.
• To provide a generalizable framework for mechanistic inference and model discrimination in spatial single-cell analysis.

Main Methods:

• TOPAZ utilizes persistent homology from TDA to quantify spatial features of cell trajectories.
• It integrates topological information with parameter inference using ABC.
• Bayesian model selection, employing the Bayesian Information Criterion, is used for model comparison.

Main Results:

• TOPAZ was validated using simulations of collective fibroblast movement.
• The pipeline accurately recovered model parameters.
• TOPAZ successfully distinguished between a baseline ABM and an extended model with alignment interactions.

Conclusions:

• TOPAZ offers a powerful computational framework for mechanistic inference from spatiotemporal single-cell data.
• The integration of TDA and ABM parameter inference facilitates model discrimination.
• The open-source availability of TOPAZ promotes its application in spatial single-cell analysis.