I am processing a dataset that contains a mix of intact icosahedral capsids, and capsids that exhibit broken symmetry due to ligand binding (they clearly look misshapen on the micrographs, as well as in the 2D classes). The intact capsids are a small minority (~10%) of the whole particle set, and I selected them from iterative 2D classification and did a follow up ab-inito reconstruction (C1 symmetry) and homogenous refinement (I symmetry), from which I now have a good volume (~3.6A).
I have been having a difficult time following the same workflow with the broken capsids, which I suspect is due to the large variation in structural heterogeneity in that group (the capsids don’t break the same way, and there is a wide variation in the intermediate structures produced as a result). Ab-initio keeps producing “squashed” volumes that are moon/donut shaped (an issue I have seen get brought up in other posts in this forum). What I do notice however, is that when I extract particles from these broken capsids with a smaller box size to investigate how the asymmetric units look (hexamers, pentamers) - they look similar to those of the intact capsids. I am therefore thinking of strategies to conduct focused refinement on these “building blocks” - the hexamers, to start – using the volume from my initial homogenous refinement (of the intact capsids) as a template. I have made a mask encompassing a hexamer from this volume on ChimeraX. I am most interested in the broken capsids, because that is where I am most likely to see density for my ligand of interest. My questions regarding this workflow are as follows:
A) Is it possible to conduct focused/local refinement using a particle stack that was not involved in the making of the input volume and its associated mask? I ask this since I am not focusing on a particular region of THAT particular volume, per se, but from a particle stack obtained from the broken capsids. If possible, what parameters should I be looking out for that will influence the outcome the most?
B) Is there any utility to the 3VDA tool in tackling the wide structural heterogeneity in my broken capsids, especially given that my ab-initio jobs with them keep failing (with C1 symmetry)? I haven’t tried using this feature yet, and am not sure what sort of progress I need to have made with ab-initio and refinement before getting to this stage.
If anyone has experience with a similar dataset, or any general tips and tricks, that would be much appreciated!