There are a few helpful things to remember: First, absolute values of global reciprocal space metrics, like R and R free, are not informative about local real space errors. So the R-value number does only inform you whether a model is perhaps complete rubbish or rather not. It also informs you about the sad fact that our models (say R ~20) are quite poor compared to data quality (say R~5). There is obviously plenty of room left for a more accurate parameterization of reality. In other words, all models are wrong but still can be useful.
Second, a real space deviation (wrong or no model, ligand) attracts one's attention, but is most of the time only a small fraction of otherwise correct model, which explains the uninformative nature of absolute values of global reciprocal space measures. Relatively, you do see changes in Rs with improvements/changes of the models, but they are small and of arguable significance. The real space inspection is of course important when the local detail (such as a ligand, or residue conformation) is the key issue for a claim or hypothesis. All the reciprocal space R-value promise goes poof if the local real space metric (say RSCC or RSR) for the item of interest says nijet. Third, the Central Limit Theorem informs us that whatever mathematically non-pathological parent distribution we sample, half or the sample distribution will be below the sample mean (average) of whatever we sample for. In other words, outliers are normal. Creepy waters? Show me one model that does not have them. The question is, does it matter? In principle, it always does. Because the structure factor is a function of each and all atoms, every correct atom necessarily improves the entire model (a tiny bit) and every poorly modeled one decreases overall model quality. Should we get bent out of shape because of this fact? The answer is it depends: Nobody truly cares about an aberrant solvent molecule, but if density does not support a ligand, a hypothesis based on its presence might suffer. Not only AI can hallucinate. Cheers, BR On Tue, Jun 3, 2025 at 9:47 AM Kelvin Lau < [email protected]> wrote: > I agree. We must remember, people are human, they have lives and can and > most likely make mistakes. > > Many metrics are also open to interpretation, even the PDB shows a quality > and percentile rank (relative to all and close resolutions), but it doesn’t > tell anyone if its "bad" or "good" that is a human judgement call. > > The PDB is not there to decide what is good and what is bad, just a > depository that allows discussions like this to occur. > > -- > Kelvin Lau > https://people.epfl.ch/kelvin.lau <https://people.epfl.ch/kelvin.lau> > > Scientist - Protein production and structure core facility – PTPSP > Co-President – EPFL Scientific Staff Association – ELSA > > EPFL SV PTECH PTPSP > AI 2146 (Bâtiment AI) > Station 19 > CH-1015 Lausanne > Switzerland > Email: [email protected] <mailto:[email protected]> > Phone: +41 21 69 30267 <tel:+41%2021%2069%C2%A030267> > If unreachable: +41 21 69 34494 <tel:+41%2021%2069%C2%A030267> > > > > On 02.06.2025, 17:14, "CCP4 bulletin board on behalf of Phil Jeffrey" < > [email protected] <mailto:[email protected]> on behalf of > [email protected] <mailto: > [email protected]>> wrote: > > > I really don't think that 22% R-free AT 2.0 Å constitutes "gross > disagreement" barring deliberate data manipulation, although it doesn't > eliminate the possibility of significant errors in lower quantities. > The review-quality validation report should report bad fits to density > (although somewhat exaggerated at time) but I don't think it has an > assay of difference density peaks (say >5 sigma) that were not fit. > Perhaps a coot-like list of unmodelled blobs would be a worthy addition. > > > I doubt the validation protocol would find, e.g., mis-assignment unless > it happened at a large scale, and the areas that it's more likely to > occur (partly disordered) aren't going to show a strong signal. Hard to > replace the human brain when trying to interpret ambiguous regions. > > > Phil > Princeton > > > On 6/2/25 9:05 AM, Oganesyan, Vaheh wrote: > > Hello, > > > > All suggestions on what and how to handle this case are good. However, > > aren’t we missing a major point? > > > > How could a model that exhibits gross disagreement with data be allowed > > in PDB? Isn’t validation protocol created to prevent cases like this? > > > > Vaheh > > > > *From:*CCP4 bulletin board <[email protected] <mailto: > [email protected]>> *On Behalf Of *Nikolas > > *Sent:* Friday, May 30, 2025 5:31 PM > > *To:* [email protected] <mailto:[email protected]> > > *Subject:* [ccp4bb] How to address deposited structures poorly modeled? > > > > *enters the BB and kneels in reverence* > > > > Dear BB, > > > > Recently, delving among papers and structures for a current project, I > > came across a deposited structure of interest that is terribly modeled. > > In the PDB resolution is 2.00 Å (R/Rfree=0.19/0.22) and the quality > > indicators are not terrible but upon opening it and checking the maps > > there are some severe mistakes in the modeling. > > > > For instance, a full chain that is in the negative Fo-Fc while positive > > density is present, some density that is quite clear for SO4 molecules > > (buffer) modeled as water and density for water molecules neglected. > > Some parts seem like “coot:runwater” has been used and not checked. The > > catalytic site is well defined. > > > > Now, I’ve started my journey in crystallography not too long ago and > > I’ve got many years, structures to solve and things to learn, and I know > > that sometimes the dataset “is what it is” but these seems quite big > > mistakes. Especially because I’ve worked with such proteins before and I > > know the systems. > > > > My question thus is: would it be right to contact the author of this > > structure and point to these issues? What’s the best way to address such > > things, possibly without coming off as obnoxious? > > > > Thank you for your wisdom. > > > > Cheers, > > > > Nikolas > > > > *bows and leaves the BB* > > > > ------------------------------------------------------------------------ > > > > To unsubscribe from the CCP4BB list, click the following link: > > https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1 < > https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1> > > <https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1> < > https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1>> > > > > ------------------------------------------------------------------------ > > > > *Confidentiality Notice: *This message is private and may contain > > confidential and proprietary information. If you have received this > > message in error, please notify us and remove it from your system and > > note that you must not copy, distribute or take any action in reliance > > on it. Any unauthorized use or disclosure of the contents of this > > message is not permitted and may be unlawful. > > > > > > ------------------------------------------------------------------------ > > > > To unsubscribe from the CCP4BB list, click the following link: > > https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1 < > https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1> > > <https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1> < > https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1>> > > > > > ######################################################################## > > > To unsubscribe from the CCP4BB list, click the following link: > https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1 < > https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1> > > > This message was issued to members of http://www.jiscmail.ac.uk/CCP4BB < > http://www.jiscmail.ac.uk/CCP4BB>, a mailing list hosted by > http://www.jiscmail.ac.uk/ <http://www.jiscmail.ac.uk/>, terms & > conditions are available at https://www.jiscmail.ac.uk/policyandsecurity/ > <https://www.jiscmail.ac.uk/policyandsecurity/> > > > > > ######################################################################## > > 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/ > ######################################################################## 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/
