This is just cool. Green Fluorescent Proteins are hugely used in biology assays including microscopy. Apparently these proteins can become magnetically responsive in the presence of a suitable cofactor. Taking advantage of this, this group used directed evolution to produce a protein with much higher magnetoresponsiveness (got brighter in a magnetic field); this could be the start of being able to switch proteins on/off with a magnetic signal, that might be less invasive than something like optogenetics.
I'm sure the directed evolution dataset will remain proprietary but ML could probably help provide clues on what changes might be necessary to make a protein 'magnetoresponsive' and help to design new ones (the Baker lab is probably already working on this).
https://andrewgyork.github.io/gfp_magnetofluorescence/#MagLOV_update