We present a novel method that utilizes generative adversarial networks to model biological systems, while constrained by biological principles. We draw inspiration from Physics-Informed Generative Adversarial Networks (PI-GANs) and extend this idea to deterministic biological models that are governed by partial differential equations. Prior knowledge of the model is encoded directly into the loss function during training by minimizing an additional biological loss term. Our method possesses the benefit of being able to learn a model through data-driven techniques, while also ensuring consistency with the laws governing the system. Additionally, it has the capability to make accurate extrapolations beyond the available data, as well as to handle noisy or incomplete data. We consider the specific case of an SIR disease model to demonstrate our results and compare the performance of our method against a vanilla GAN.

This content is only available as a PDF.
This is an open-access article distributed under the terms of the Creative Commons Attribution 4.0 International License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. For a full description of the license, please visit https://creativecommons.org/licenses/by/4.0/legalcode.