Density became worse after NU refinement

Hello, everyone! I am a rookie to Cryo-EM data processing. I encountered a problem that I got a good class of particles after Ab-Initio Reconstruction but the density became worse after NU refinement. The parameters for NU refinement were default except that the Initial lowpass resolution (A) was changed to 10 Å. Any ideas about that?

the map colored by yellow is the initial model and the gray one is the map after NU refinement.

Hi @Verdandy,

This looks like a very small molecule. In such cases, the dynamic masking procedure in NU-refine often makes things worse - I would suggest switching it off (by setting the start resolution for dynamic masking to 1Å).

Alternatively, you can try initiating local refinement starting from the poses you have generated with ab initio.


1 Like

Hi @olibclarke ,
Thank you for your suggestions. I have tried to set the start resolution for dynamic masking to 1Å during NU-refinement but got similar result as before. I also tried local refinement using the initial model and corresponding particles without any improvement on map density. I wonder whether the strong micelle signal will hamper the resolution and how to solve this problem?

this is the result after NU-refinement

this is the result after local refinement

Hmm - how many particles do you have remaining? It may also help to try to clean up your particle set more, e.g. by heterogeneous refinement with one “good” model and several junk classes

Well, there are only ~45,000 particles for this good class.

Ah I see. What molecular weight is the sample? It looks rather small?

the molecular weight of monomer is ~70 KDa and the protein is a dimer.

Hmm - the density shown looks much smaller than 140kDa - maybe some domains disordered? If the protein is a symmetric dimer, have you tried enforcing C2 symmetry in either ab initio or refinement?

Not yet, I will try it later. BTW, it is low the resolution that makes it look like small?

No, it looks small in terms of the number of structural elements present. Looks like the equivalent of ~2x7TM bundles, with nothing outside the membrane.

Hi Oli,
I have met the same question with Verdandy AND I just tried your method which is initiating local refinement starting from the poses you have generated with ab initio.. However, it always failed and showed I should put some half sizes. I try two kinds of local refinement, but both failed. So, is anything wrong with my operation ? Thank you.

Hi Qu,

You need to set “force GS-split” to true (it is off by default):


Thanks, I will try it immediately.

It seem still failed :joy:

Hi Oli
I guess I get your point. The local refinement must be done after Homo/heter refinement. So the mask should be put from the homo refinement. Thanks a lot.

Yes, you need to provide a mask - doesn’t have to be from refinement, you can make it from the ab initio model or however you like

1 Like