Choosing Reference Map and Mask in 3D Auto Refine After CTF Refinement and Bayesian Polishing

[moranchu]

I have a question regarding the choice of reference map and reference mask in the 3D auto refine step after CTF refinement and Bayesian polishing, as outlined in the Relion tutorial.
Question:

1. CTF Refinement Followed by 3D Auto Refine:

In the tutorial, after performing CTF refinement, the 3D auto refine step uses the full run_class001.mrc as the reference map without applying a reference mask. Why is this the case?

2. Bayesian Polishing Followed by 3D Auto Refine:

After Bayesian polishing, the tutorial suggests using one of the half-maps as the reference map and applying a reference mask in the subsequent 3D auto refine step. What is the rationale behind this difference?

3. Sequential Optimization:

If I need to repeat these optimization processes sequentially (i.e., perform CTF refinement, 3D auto refine, postprocessing, Bayesian polishing, and then another round of 3D auto refine and postprocessing,), how should I choose the reference map and reference mask for the intermediate and final 3D auto refine step?

Thank you for your help!

[biochem-fan]

Are you talking about SPA or STA?

If you mean SPA:
(I cannot comment on STA because I have zero experience with it)

CTF Refinement Followed by 3D Auto Refine:
In the tutorial, after performing CTF refinement, the 3D auto refine step uses the full run_class001.mrc as the reference map without applying a reference mask. Why is this the case?

You should use a mask; if the tutorial does not, we should fix it.

Bayesian Polishing Followed by 3D Auto Refine:
After Bayesian polishing, the tutorial suggests using one of the half-maps as the reference map and applying a reference mask in the subsequent 3D auto refine step. What is the rationale behind this difference?

By starting from half maps, you can use weaker low pass filtering and save refinement time. But you can always safely use "run_class001.mrc".

[moranchu]

Are you talking about SPA or STA?

If you mean SPA: (I cannot comment on STA because I have zero experience with it)

CTF Refinement Followed by 3D Auto Refine:
In the tutorial, after performing CTF refinement, the 3D auto refine step uses the full run_class001.mrc as the reference map without applying a reference mask. Why is this the case?

You should use a mask; if the tutorial does not, we should fix it.

Bayesian Polishing Followed by 3D Auto Refine:
After Bayesian polishing, the tutorial suggests using one of the half-maps as the reference map and applying a reference mask in the subsequent 3D auto refine step. What is the rationale behind this difference?

By starting from half maps, you can use weaker low pass filtering and save refinement time. But you can always safely use "run_class001.mrc".

SPA it is. Thank you!