Morphogenesis as Inference
We recast developmental pattern formation as approximate Bayesian inference over a generative model of tissue. The resulting dynamics reproduce classical Turing patterns, recover experimentally observed stripe-spot transitions in zebrafish, and predict a previously unreported bistable regime in chick feather primordia.
Reframing
A developing embryo is not executing a pre-written program; it is *inferring* its own shape from noisy positional cues. We make that intuition precise.
Model
Tissue is a Markov random field; positional information is a noisy projection of a latent target morphology. Patterning dynamics minimise variational free energy in this field.
Result
The same equations reproduce Turing patterns, Hopfield-style stored morphologies, and a previously unreported bistable regime — feather spots that admit two stable spacings depending on initial conditions. We observe both experimentally in chick primordia cultures.
Why it matters
It reframes developmental biology as a problem in statistical mechanics, and makes concrete predictions that can be tested with current experimental tools.