Refinement is adding artifacts


I am having issues with both Homogenous and Non-Uniform refinements. Ab-initio map looks good but refinement maps (attached)
have a extra layer on it. The jobs were ran with the default parameter. Is there a parameter which should be changed or a way to fix this issue?

Thanks in advance,

would love to see a great explanation from others - but empirically, this happens when ~half the particles are well-assigned, and ~half the particles are junk which do not fit the model. I think the best way forward is a combo of Het refinement, multi-class Ab Initio, or (least favorite) selecting 2D which represent 1 structure from different angles.

Agree. Including “junk” being actually alternative conformations due to flexibility. In our case further 3D and 3D-based classification helped a lot. Of course you need to have plenty of particles at that stage to move further. I’d be interested to see what 3DFlex does in cases like this, with the same particle set. You might need to try different solvent masks.

Thanks for the suggestions.
I did two rounds of 2D classification, 400K particles were subjected to an ab initio reconstruction with five classes, no similarity, C1 symmetry, and initial alignment resolution of 12 Å. Good particles from classes were directed into a second ab initio reconstruction with three classes, no similarity, C1 symmetry, and 7 Å initial resolution. The 2 best-resolved class possibly the two population of particles, together with the worst-resolved class from the first ab initio reconstruction, were used as references for 2 rounds of heterogeneous refinement, together with the particle set selected from the first round of ab initio reconstruction. After heterogeneous refinement, where the initial resolution of the input volumes was set to 7 Å, 2 classes with 200k particles each in a non-uniform (NU) refinement. After 2 rounds of NU refinement, I get a map but it has some weird features at the periphery, which doesn’t look like part of the map. Can someone help me how to get rid of it?
Also, is there a way to stop the dynamic masking?
What are the criteria or allowed changes for the dynamic mask near and far parameters?
Attaching here the map of NU refinement.

I will try 3D flex in the later steps and will be happy to update you with results.