I have a good class that I want to subclassify, but when I attempt to subclassify it with a 2 class heterogeneous refinement, the resulting two classes represent two portions of the orientation distribution, rather than distinct conformations.
I have seen similar behavior several times with cryosparc, but never to the same extent with relion.
See attached - this was with an RyR dataset - has anyone seen the same behavior and found a remedy?
Happy to see that I am not the only one!
It happens very frequently in Relion too.
Regarding this it would be nice if cryosparc showed the orientation distributions of each class during heterogeneous refinement, like it does in homogeneous refinement and ab initio reconstruction. It helps to identify the issue at a glance (whereas it may not always be obvious looking at the map central slices)
So I found one thing that seems to help here, at least in this case. The heterogeneous refinement was initialized with two identical structures, and I end up with the distorted structures shown above. If I iterate the heterogeneous refinement a couple of times - taking the output structures from the first round and using them as input for another round of heterogenous refinement, with the same particle set, it seems to help a lot - I have no idea why.