Trainable Morphogenesis Simulation in JAX.
morphix is a python/JAX package to simulate and train morphogenesis, i.e.,
the development of organsisms from a single cell. morphix simulates
- mechanical forces between cells,
- secretion/diffusion/sensing of molecules between cells,
- cell-internal reactions, and
- cell division events.
The functions underlying all simulated components (except for the mechanical forces) are learnable through a combination of auto-differentiation in JAX and reinforcement learning.
In the future, this will allow learning the development rules that give rise to different morphologies.
frames_2026-02-26_11.44.31.mov
This package is under active development (pre-release) and the API is subject to change.
This package was developed in an independent effort to Jax-Morph and shares a lot of its goals. Please refer to the publication Engineering morphogenesis of cell clusters with differentiable programming to learn more about Jax-Morph and what it can do.
At the moment, Jax-Morph is a more mature package for morphogenesis simulation, highly customizable, and has already been used to uncover principles of development.
To deploy a new version, first make sure to bump the version string in
morphix/__init__.py. Then create an annotated tag, and push it to github.
This will trigger the deploy.yaml workflow to upload to PyPI
git tag -a vX.Y.Z -m vX.Y.Z
git push upstream --follow-tags