@olibclarke Hmm, this is a good question for sure. Naively, I would suggest that experimenting with different mask tightness would probably be the best approach, and this can be informed by how much flexibility you suspect (e.g. if you are only looking for small scale motion, then tighter masks like the ones outputted from refinement may be okay). I haven’t experimented enough with mask tightness to give much insight on this unfortunately – possibly @apunjani could give more information than I can.
@adtaheri Running a consensus refinement with all particles & a mask that includes the whole 330 kda complex (plus the entirety of the flexible region) may be the best option – this way, you end up with all particles aligned to the same common reference, which is important as otherwise it’s possible that 3D variability could end up detecting small differences in rotation between each class, rather than the desired movement. This consensus refinement might not reach quite as high resolution as any of the individual 3D classifications, but for the purposes of 3D Variability, this is the approach we recommended in our tutorial.