Training rules for local, non-genetic learning in phase-separating systems

Determine explicit local training rules that enable phase-separating molecular systems to learn non-genetically by updating effective interactions through post-translational modifications and modulation of hidden-component expression levels.

Background

Physical learning suggests that materials can acquire functions through exposure to examples, but a mechanistic basis for locally updating interaction parameters without genetic change is needed. Post-translational modifications and hidden-component concentration changes are proposed levers that reprogram phase behavior.

Formulating concrete, local training rules would establish how condensate-based systems could adapt computationally on cellular timescales.