[phenixbb] Re: ADP refinement in a CryoEM map of heterogeneous quality

Hi Pavel, Thank you for this clear answer. I don't know why my first message did not reach you (it is in the archives but I can put it very simply): I was surprised by the low B-factors in a mobile region of my CryoEM structure at 2.6 angstroms FSC (see figure attached, also obtainable by https://filesender.renater.fr/?s=download&token=9688951f-830a-4cb4-a09f-b73ffb86620f in case the BB doesn’t like attachments). It never occurred to me that the problem could come from refining against a sharpened map (I leave the unsharpened for the very last moment before deposition...). After your reply, I tried the unsharpened map. It partially solved the problem, indeed (the new range starts at ~32 angstroms square, which sounds compatible with our resolution range, CC_mask is 0.82), but the left side of the light chain is still too cold. We are specialized in myosins, I've seen many, many structures of them and this is the first time I see lower B-factors for the light chain than for the lever arm helix. This also does not correlate with the map quality of that region, so I am still puzzled. I got your message. I'm OK for keeping those Bs in the corresponding column anyways. And sure I can ship you the maps and model in private, if you think it could be useful. All the best! --------------------------------------------------- Carlos KIKUTI, PhD Structural Motility Team UMR144 - CNRS - Institut Curie 12 rue Lhomond - 75005 Paris, France [email protected] mailto:[email protected] Le 07/03/2026 23:50, « Pavel Afonine » mailto:[email protected]> a écrit : Hi, from “just an update,” I suspect there was an original message that I haven’t seen (not in spam either). From what I read below: - “very low values” for B need to be defined; - refined B around zero likely means you refined against a sharpened map and can’t be “fixed” unless you use the original (unsharpened) map; - changing the context of a file with a refined model generally invalidates the statistics and is not advisable; - ADP, or B factors, have a well-defined physical meaning, and that is what the corresponding column in your model file is reserved for. Setting it to something else sounds nonsensical to me, unless you want to use it to color your model by whatever values you place there and then throw the model away; - since in cryo-EM there is nothing like Rfree in MX, you need to manage overfitting (which is much less of a problem in cryo-EM than in MX) by watching CCmask and overall geometry metrics. Sometimes decreasing CCmask can mean “lowering Rfree,” essentially, which is not a bad thing of that happens at the cost of geometry improvement. I can provide more diagnostics if you send the model and map. Good luck! Pavel On 3/4/26 07:35, [email protected] mailto:[email protected] wrote:

OK, just an update:

My cheating did not work. Even after I changed the chain name to get everything as chain A, and set up the ADPs of the light chain to 140, they went back to very low values, making the CC_mask go worse (although still > 0.7). I also tried increasing the number of iterations to 600, no success. I'm considering placing Q-scores in the ADP column, as I believe that would be more realistic. Any strong opinions about that? _______________________________________________ phenixbb mailing list -- [email protected] mailto:[email protected] To unsubscribe send an email to [email protected] mailto:[email protected] Unsubscribe: phenixbb-leave@%(host_name)s