Modelling proteins with a self-organizing differentiable system

TLDR: Physics inspired protein modeling effort incorporating self assembly as an inductive bias. Com- bining ML and molecular dynamics — except it is not trained on dynamics, instead just static structures from the PDB, and you potentially get dynamics for free. Other perks: interpretability, computationally tractable to model larger proteins

Modelling proteins with a self-organzing differentiable system.pdf183.6KB