Hi All I think this discussion is important, and is best thought about through the frame of Cryo-EM, because we image in real-space and essentially measure our phases we routinely resolve very diffuse regions where the chain is present but extremely mobile, this gives the temptation to build highly complete models with regions that would be much more uninterpretable in an X ray experiment (solvent masking/phase error etc..).
In X-ray its rarer and the range in local resolution/map detail is going to be a bit more constrained by the requirements of what's possible in a viable diffracting crystal, but its still fairly common in high-solvent content crystals with diffuse packing etc.. Usually, these regions can be built in many different ways, and all of the possible conformers are usually more accurate relative to alphafold purely in terms of average distance between an atomic coordinate and its "true" position (if we assume the "true" position is in the density somewhere". You might also think that a flexible region built into diffuse density which is geometrically favourable is likely to be present some of the time. However, if we cannot even locate broad features that allow us to (fairly) accurately infer the positions of other atoms (i.e place bulky sidechains to collapse the conformations into a more useful range of possibilities), then the improvement in the spacial accuracy is could be considered meaningless, and gives the illusion of certainty (Just with high level or thermal motion or whatever) to an uninformed interpreter. The ambiguity can sometimes be anisotropic because of the shape of known features such as a glycan tree where the overal direction can be inferred very easily but you can rotate each unit without much cost to RSC/geometry. In truth, the best fit to density that is an average of many conformers could actually result in a conformer that almost never occurs! In other words you cannot just consider a 6-8A region of an EM map (or similar in X-ray) as simply a dilated/blurred version of a 3A region because we don't know how the various conformers contribute to the whole. The obvious example is where the sidechains of two conformers flip over and produce density in between the two true main-chain locations that resembles a false position of the main chain when combined, In EM flexible regions can be averaged over high point group symmetry to produce very sharp features that can look very similar to nice near-atomic features. Some kind of standardised score would be really useful perhaps something connected with the uniqueness of the RSC peak as conformational space is sampled by MD (computationally expensive I'm sure š) and how steep the RSC peak is with some weighting by strain/geometry. But I don't think any technique can really tell us when we are better off replacing an experimentally determined (semi-determined/semi-inferred) feature with a computationally predicted feature, especially now that the best predictions are done by deep-inference/diffusion rather than by simulating the physical chemistry of the molecule. Best Matthew. ________________________________ From: CCP4 bulletin board <CCP4BB@JISCMAIL.AC.UK> on behalf of Tim <00008ec2c4efb208-dmarc-requ...@jiscmail.ac.uk> Sent: 04 March 2025 09:06 To: CCP4BB@JISCMAIL.AC.UK <CCP4BB@JISCMAIL.AC.UK> Subject: Re: [ccp4bb] IDS in PDB Dear Pavel and Oliverio, In our group we have had recent discussion about this issue, and I second Pavel's suggestion to introduce such a confidence measure. When interpreting cryo-EM (but also X-ray crystallography) maps with models, we often face the problem that we have complete models (thanks to Alphafold) but incomplete density. Most often the lack of density is due to conformational flexibility. So the question is what to do with those parts that lack map support? Some favor the option not to model these parts. However, usually we are pretty certain that these parts are present, but too flexible to be observed. So I personally think that structural models should be deposited as complete as possible. Having such a confidence measure would facilitate the interpretation of our structural models, also because it seems that not many actually open the associated deposited maps/densities when interpreting deposited structures. Best wishes, Tim On 2025-03-04 00:27, Pavel Afonine wrote: Greetings, It's hard to disagree with this! Resolution, occupancies, and B factors only indirectly suggest what's visible and what isn't ā and they can be especially difficult to interpret correctly for non-specialists. Perhaps a local confidence measure ā similar to pLDDT for predicted models ā could address this by condensing into a single number everything we know about the model quality and how well it fits the data, computed per atom or per residue. All the best, Pavel On Mon, Mar 3, 2025 at 7:21āÆAM Italo Carugo Oliviero <olivieroitalo.car...@unipv.it<mailto:olivieroitalo.car...@unipv.it>> wrote: A brief reflection on IDPs Increasingly, people with a computer science background are analyzing the data deposited in the Protein Data Bank. In the case of conformation disorder analyses, they consider residues that are explicitly stated to be disordered (the old REMAR 465 records). This is not quite correct as there are two problems. - The first is that some crystallographers consider āvisible,ā and deposit their coordinates, even amino acids that have stratospheric B-factors, so large as to indicate that those amino acids are definitely āinvisibleā in electronic density maps. - The second problem has to do with crystallographic resolution. The amount of āinvisibleā amino acids increases as the crystallographic resolution decreases. At low resolution, electron density maps are often not very detailed, and some parts of them cannot be interpreted. But this does not mean that the amino acids found there are definitely āinvisible.ā It simply means that resolution might be insufficient. Editors and reviewers may find it useful to keep these considerations in mind when evaluating articles on conformational disorder submitted by scientists that lack a structural biology background. Or is there something else that can be done? ________________________________ To unsubscribe from the CCP4BB list, click the following link: https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1 [jiscmail.ac.uk]<https://urldefense.com/v3/__https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1__;!!PDiH4ENfjr2_Jw!EtK2wXQtXD6V5IkAN2tFANSllLDc3GvyUcTOuMsQ6hkdjet5Lv96sa2FttirWNefYcQTWUrEusAtwENbWI-uyCTIKUkSLTYduPH1kUpYCVPl3w$> ________________________________ To unsubscribe from the CCP4BB list, click the following link: https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1 [jiscmail.ac.uk]<https://urldefense.com/v3/__https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1__;!!PDiH4ENfjr2_Jw!EtK2wXQtXD6V5IkAN2tFANSllLDc3GvyUcTOuMsQ6hkdjet5Lv96sa2FttirWNefYcQTWUrEusAtwENbWI-uyCTIKUkSLTYduPH1kUpYCVPl3w$> ________________________________ To unsubscribe from the CCP4BB list, click the following link: https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1 [jiscmail.ac.uk]<https://urldefense.com/v3/__https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1__;!!PDiH4ENfjr2_Jw!EtK2wXQtXD6V5IkAN2tFANSllLDc3GvyUcTOuMsQ6hkdjet5Lv96sa2FttirWNefYcQTWUrEusAtwENbWI-uyCTIKUkSLTYduPH1kUpYCVPl3w$> ######################################################################## To unsubscribe from the CCP4BB list, click the following link: https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1 This message was issued to members of www.jiscmail.ac.uk/CCP4BB, a mailing list hosted by www.jiscmail.ac.uk, terms & conditions are available at https://www.jiscmail.ac.uk/policyandsecurity/
