I am running Non uniform refinement with the default settings, 100K particles (output of heterogeneous refinement job), a volume input (~7 Angstrom output of a Heterogeneous refinement job), and no static mask input. The job runs, but then finishes after 3 iterations, and the corrected mask in the FSC plunges down vertically to 0 at around 6-7 Angstroms. The output is a low quality map, when I would expect something closer to the 3 Angstrom range. This happens regardless of what volume input I use. Any thoughts? Is this an issue with the dynamic masking? I have never needed to input a static mask for this job in the past. Im happy to provide additional information to clarify my situation if it would help! Here are the GSFSC outputs
Did you confirm that this particular dataset can in fact produce a 3 Å resolution model? Has the gain reference been correctly applied during movie import?
for anyone that may be experiencing this, I believe it is due to unwanted/junk particles (at least it was in my case).
I had the exact same scenario of that “plunging” FSC curve, in addition to NU 3D refine stalling at ~110k particles, even though I fed it ~500k particles. It eventually would process all 500k, but the plots and maps were awful.
Not sure if the original post was from a small SNR protein, but I think I was too liberal in picking even my 3rd round of 2D classes from template picking for my small (~100 kDa) protein.
