Local Refinement

Hello everyone!
Someone could explain or send some documentation about Local Refinement?
I’d like to understand the difference in the algorithm and if I can use it with the same mask (including almost the whole protein complex) that I used in NU-refinement.
I did this and the resolution improved, but I don’t know if it’s a good thing to do.

I’m learning, so any information will be helpful!

Hi @deborahcm, here are a couple of resources that might be helpful:

Local refinement post

Local refinement case study

1 Like

If you are getting an improvement from some additional local search iterations, you can also forcing some extra iterations in the NU refine itself (“force extra iterations”). Setting the FSC threshold to 0.75 instead of 0.5 can also help.


you mean the FSC threshold for the local filtering, right? I wish it was possible to set the FSC threshold used for overall alignment, would help to mitigate overfitting for small things I should think

1 Like

That makes sense too. I’ve found changing the FSC for the local filtering to 0.75 to effectively remove any residual overfitting and give a really nice, smooth micelle. Unfortunately I couldn’t find a good AWF change to get the same outcome as easily.

Another approach for the overall alignments would to be able to specify a maximum alignment resolution, but use the lower resolution value of that and the FSC in the last iteration. That way, the global alignments in the initial iterations will work well, but you could still have a hard limit at the end.

1 Like

Yeah it definitely helps, I still have cases where it doesn’t completely fix the problem though… and unfortunately you can’t change those params in local refinement with NU-refine (and local refinement also ignores any beam tilt etc refined in global CTF refinement)

1 Like