Multiple conformation dataset prosessing

Cryo-EM Data Processing
1

Hi all,

I have a dataset of Spike combine with Fab. When I processing the dataset, I find that there are more than 3 kinds of comformation. However, in every conformation the particles numbers is very low about 20,000 particles. So it caused the RBD and fab resolution is not good, but spike NTD could got 3.7 A. Is there any sugesstions how to improve the resolution of RBD and Fab?

2

Local refinement with a focussed mask?

Achieving high resolution for RBD/FAB is going to be tricky with such a low number of particles and relatively low NTD resolution, however.

3

Hi Jason, I am assuming that the conformational heterogeneity results from interdomain flexibility between Spike RBD and NTD; to the best of my knowledge, a typical Fab is a tight binder for which the PPI with antigen is usually conformationally homogenous.

If this is correct for your case, I would do a particle subtraction and local refinement using a local mask that covers RBD and the variable domains (VH and VL) of the Fab; since the elbow region of Fabs are often somewhat flexible, I would exclude the constant domains (CH1 and CL) from the local mask.

Hope that helps :slight_smile:

4

@rbs_sci and @cquan

Thank you for the suggestion. We did try local refinement with a focused mask on the RBD/Fab region, but due to the limited number of particles and the flexibility in the NTD area, the resolution improvement was minimal.

We are considering further data collection to increase the particle count and improve the overall map quality

1 Like
5

Hi Cquan, Thanks for your suggestions. I have tried the particle subtraction job, focusing on RBD and Fab. Finally, I got a ~4A map with 500k particles, but this resolution is hard to clearly confirm the secondary structure of Fab and RBD. Do you have any experience improving the resolution after particle subtraction? :smile:

6

Hi Jason, IMO 500K particles is a fairly large number; assuming that these particles are from a Krios collection and have been cleaned by 2D classification, I would expect a resolution substantially higher than 4A.

Here is what I am thinking about:

  1. How is the orientation distribution? Resolution would take a hit in presence of preferred orientation. If that is the case, 2D rebalance might help.
  2. Is there substantial conformational flexibility at the elbow region of the Fab? To the best of my knowledge, some germline alleles have more flexible elbows than others. If this is the case for your Fab, I would use a local mask that covers only the variable region (VH + VL) and RBD (assuming that RBD is conformationally rigid) and subtracting the signals corresponding to the constant region (CH1 + CL) and NTD.
  3. What is the affinity of your Fab? My experience is that weaker binders may result in conformational heterogeneity at the PPI. If that is the case, you may need to do 3D classification to find the best conformational class.
7

Hi Cquan,

Thanks for the thoughtful feedback – I really appreciate your insights. You’re absolutely right that 500K particles should, in principle, yield better than 4Å resolution, especially after 2D classification. I’ve processed the data through DeepEMhancer , and it produced a map with an estimated resolution of ~3Å. This improved map quality has been very helpful for confirming the main-chain assignments.