Hi - I am not sure how well local refinement works for flexible regions of a molecular weight around 50 kDa.
I’m working with a mask that surrounds this region, but I did not do particle subtraction. I do not need high resolution, but only to resolve secondary structure. I see that this information is there while the job runs, but it gets lost by the end of the job. Would it make sense to limit resolution to 5, or is there any other option one could try for this specific case ?
Many thanks
Jacopo
Hi @marino-j,
Thanks for the post. When you download and view the volumes from intermediate iterations, do you notice that secondary structure density gets noticeably worse over time?
This seems like an overfitting issue, so limiting the alignment resolution could help as you suggested. You may also want to check out our section on Mask troubleshooting on the local refinement job page. It highlights the importance of mask padding, and using static over dynamic masking. Also note that 50 kDa is quite a bit smaller than is recommended for local refinement unfortunately, so you may have to apply tighter priors over rotation/shift in order to regularize.
Otherwise, if the masks are adequate, this indicates that our termination criteria is not detecting convergence properly. If you are open to it, it may be of interest for our development to work with your data directly as we’d like to be able to guard against this. I can follow-up further if so.
Best,
Michael