Problem of the refiment of non-dominant conformation of small membrane proteins

Cryo-EM Data Processing
1

Hello, I was doing refinement of a small membrane protein(~80kD). I successfully got the outward conformation of this protein with 3.4 angstrom. However, when I looked into 2D class clearly, I believed there exists another conformation. So I used the same family of this protein which is inward state to isolate inward open particles. And I got 500k particles. After ab initio, whether I did local refinement or NU refine of initial model, I saw a clear shake of FSC curve:

J1477_fsc_iteration_003
J1477_fsc_iteration_003506×293 25 KB

But the density map looks not bad
With so many particles remained, I believe the final resolution could be 3.7 angstrom at least. However, this inward map can not successfully isolate inward particle when doing hetero refine with outward map and bad map as input. Besides that, the. NU and local can not improve resolution with first round is 4.5A and final is about 4.3A. When I did 3D classification, things got worse: FSC curve not correct and the map is distorted:
How can I do to get high resolution of this inward state?

2

The map after NU refine based on ab

screenshot-1748363228421
screenshot-17483632284211071×901 78.6 KB

3

3D classification result:

screenshot-1748363538389
screenshot-17483635383892779×199 232 KB

4

I would be very wary of this approach (using the structure of a related protein as a template for baited classification). How exactly did you do it?

The reason why I would be wary is that it is very vulnerable to template bias. If you have enough of this other state that you can see it in 2D classification, you should be able (usually) to isolate it by careful 3D classification after a consensus refinement.

Cheers
Oli

1 Like
5

Assuming that you don’t have serious issues with preferred orientations, it would be nice to test 3DVA in this case, filtering to 12 angstroms or so, to see if it can distinguish the open particles. You can be less stringent on the picking (blob picking or a very low-passed template) to maximize the number of initial particles, just make sure the set is clean of junk particles.