thank you to all my co-authors and contributors to our open source code base! (and thanks @miroastore.bsky.social for making this movie)

a toy demonstration: in only a few blocks of code, we can use cryoJAX and JAX automatic differentiation to implement a gradient-based optimization that recovers conformation A starting from conformation B. check out the code (roughly) used here: michael-0brien.github.io/cryojax/exam...

this has led me to cryo-EM and inspired my development of cryoJAX. i think cryo-EM data has great promise to inform physics models, but leveraging the data is computationally burdensome and a challenge of high-dimensional statistical inference.

some background: in my phd, i'm working towards bridging protein and cellular scale behaviors in microtubule self-organization. the problem is that the scale of individual proteins is extremely difficult to observe in situ.