Generative diffusion model surrogates for mechanistic agent-based biological models.

Area of Science:

• Computational biology
• Biophysics
• Artificial intelligence in biology

Background:

• Mechanistic, multicellular, agent-based models (e.g., Cellular-Potts Model, CPM) are crucial for single-cell resolution biological investigations.
• The computational expense of CPMs at large scales hinders their application and analysis.
• Stochasticity in CPMs complicates the development of accurate surrogate models.

Purpose of the Study:

• To develop an AI-powered surrogate model for accelerating CPM simulations.
• To address the challenges posed by stochasticity in surrogate model development for CPMs.
• To investigate in vitro vasculogenesis using a generative AI surrogate of a CPM.

Main Methods:

• Leveraged denoising diffusion probabilistic models (DDPMs) to train a generative AI surrogate of a CPM.
• Employed an image classifier to identify unique regions within a 2D parameter space.
• Utilized the classifier to assist in surrogate model selection and verification.

Main Results:

• The CPM surrogate model successfully generated configurations 20,000 timesteps ahead of a reference.
• Achieved an approximate 22x reduction in computational time compared to native CPM code execution.
• Demonstrated the feasibility of using DDPMs for developing digital twins of stochastic biological systems.

Conclusions:

• AI-driven surrogate models, specifically DDPMs, can significantly accelerate complex biological simulations like CPMs.
• The developed approach offers a pathway to overcome computational limitations in agent-based modeling.
• This work paves the way for creating accurate digital twins of stochastic biological systems, enhancing research capabilities.