Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers!!!!!
Once more to those who feel offended by the structures in discussion: I’d be very careful at judging low resolution structures. This is a tricky business requiring a lot more info than just the PDB validation report. The 3+ to 4 A resolution range is a particularly deceptive one: The crystallographer does not have much data given the model parameters (perhaps consulting his figure showing determinacy for coordinate refinement might help) http://www.ruppweb.org/Garland/gallery/Ch12/pages/Biomolecular_Crystallography_Fig_12-11.htm At this resolution one has about enough data to keep enthusiasm up but at the same time it is not quite yet bad enough to throw up the hands and admit that that one is de facto modelling with a few X-ray restraints (i.e. data), requiring correspondingly suitable refinement protocols (and discipline, aka mental restraints in addition to stereochemical restraints). One is easily spoiled by looking exceptional 2A structures of huge complexes, but nature (I do not mean the journal but the same time would not exclude it) is often cruel. Particularly in Molecular Replacement structures, and here particularly in those with multi- segment/domain models, there are almost always parts that fit well and others that fit poorly - with simply not enough data at the given resolution to improve the poor parts sans additional phase information. Bias issues have been discussed and need not be iterated here. Pavel is correct in pointing out that a model with better geometry is also a more plausible model. What we cannot tell sans supporting density is whether it is a more accurate model, although I have rarely seen an improvement in geometry giving worse density fit. Usually a mess remains a mess - there is (at this resolution) no free lunch. The key question is again – does the model justify the specific conclusions drawn from it? If a poor model is better than no model at all, be it, as long as this is recognized and not used as an excuse for careless work. Facile dictu, difficile factu. Best, BR For every sufficiently complex problem there is an answer that is simple, clear and wrong. LH Mencken From: Narayan Viswam [mailto:nvisw...@gmail.com] Sent: Monday, April 27, 2015 12:02 PM To: b...@hofkristallamt.org Cc: CCP4BB@jiscmail.ac.uk Subject: Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers! Sorry can't help it. The aggressive replies are mainly from senior PIs from US - friend in need is friend indeed.
Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers!!!!!
Hi Robbie and Co, These things are happening now too. Look at the entry 4x4m. The paper got published in January, PDB released coordinates in April. That means reviewers did not have a chance to look even at validation report. In my opinion, whatever it is worth, every journal dealing with crystal structures should, at the very least, request the validation report from PDB including Nature, Science and PNAS. What is also interesting that at the end PDB released the coordinates with large number of outliers. I don't think those can be justified with low resolution of the data. Regards, Vaheh Oganesyan www.medimmune.com From: CCP4 bulletin board [mailto:CCP4BB@JISCMAIL.AC.UK] On Behalf Of Robbie Joosten Sent: Friday, April 24, 2015 2:37 AM To: CCP4BB@JISCMAIL.AC.UK Subject: Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers! The PDB_REDO entry for 3bdn was pretty old, so I replaced it using a newer version of PDB_REDO that can use nucleic acid restraints from LibG: http://www.cmbi.ru.nl/pdb_redo/bd/3bdn/index.html. Obviously, the new structure model is far from brilliant (PDB_REDO doesn't rebuild at this resolution), but Molprobity seems to like it quite a bit more. I agree with the replies so far in that: - The topic starter was rather blunt and could have been more subtle. He should probably go work in the Netherlands ;) - Building structure models at 3.9A is incredibly difficult. - The tools we have now are much better than in 2008. However, we should not act like 2008 were still in the dark ages of crystallography. There are a lot of good structures available from that time (and also from long before) even at that resolution. That is not surprising seeing that we also already had very good building and refinement tools available. We also had enough validation tools available to tell us that this particular structure model isn't very good. I really believe that a good crystallographer that was not pressed for time (or at least didn't rush) could have done better with the data and the tools available. I'm now going to hide behind an asbestos wall to say this: The manuscript was submitted in July 29th 2007, the PDB entry was deposited November 15th 2007. That means that the referees probably did not have a chance to see the finished structure model, at least not in the first pass. This implies that the authors didn't want to deposit the model on time. There are a whole lot of excuses for this, that are fortunately dealt with now (http://onlinelibrary.wiley.com/doi/10.1107/S0907444913029168/abstract), but the referees could have been a bit more critical. They should have at least seen that the supplemental table 1 did not show any Ramachandran statistics. We can only speculate what happened. I'm guessing that the authors didn't finish the structure yet and rushed the publication through to avoid being scooped or for the general glory of a Nature paper. To bad that came at the expense of the crystallography. Cheers, Robbie Date: Thu, 23 Apr 2015 18:43:13 + From: kell...@janelia.hhmi.orgmailto:kell...@janelia.hhmi.org Subject: Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers! To: CCP4BB@JISCMAIL.AC.UKmailto:CCP4BB@JISCMAIL.AC.UK Is it in pdb redo? Take a look here: http://www.cmbi.ru.nl/pdb_redo/bd/3bdn/index.html JPK From: CCP4 bulletin board [mailto:CCP4BB@JISCMAIL.AC.UK] On Behalf Of Misbah ud Din Ahmad Sent: Thursday, April 23, 2015 2:28 PM To: CCP4BB@JISCMAIL.AC.UKmailto:CCP4BB@JISCMAIL.AC.UK Subject: Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers! Dear Phoebe A. Rice, I didn't mean to discredit the work but the statistics of the structure just shocked me at the first instance. I could for example point out to another structure 1ZR2, which has the same resolution (the protein Molecular weight is almost the same) and the statistics are: R-work: 0.27 Rfree: 0.319 Ramachandran outliers: 2.8% The structure was solved by a combination of MIR and MAD three years earlier in 2005 and definitely they didn't had better softwares. As per your advice, I tried one cycle of refinement of this structure in Phenix and the statistics are: Start R-work = 0.2816 ; Start R-free = 0.3551 Final R-work = 0.2887 ; Final R-free = 0.3671 Ramachandran outliers: 18.91% Rotamer Outliers: 25.19% Being a crystallographic community, isn't it our responsibility that if better softwares are available now, we try to re-refine our older structures and deposit better models in the PDB. That would help the users immensely. Misbha On Thu, Apr 23, 2015 at 7:06 PM, Mark J van Raaij mjvanra...@cnb.csic.esmailto:mjvanra...@cnb.csic.es wrote: The abstract of the papers says they used MIR. Mark J van Raaij Dpto de Estructura de Macromoleculas Centro Nacional de Biotecnologia - CSIC c/Darwin 3 E-28049 Madrid, Spain tel. (+34) 91 585 4616 http://www.cnb.csic.es/~mjvanraaij On 23 Apr 2015, at 18:57, Todd Jason Green wrote: My guess
Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers!!!!!
7 out of 1361 Ramachandran outliers (all prolines) doesn't seem high to me. From: CCP4 bulletin board [CCP4BB@JISCMAIL.AC.UK] on behalf of Oganesyan, Vaheh [oganesy...@medimmune.com] Sent: Monday, April 27, 2015 9:04 AM To: CCP4BB@JISCMAIL.AC.UK Subject: Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers! Hi Robbie and Co, These things are happening now too. Look at the entry 4x4m. The paper got published in January, PDB released coordinates in April. That means reviewers did not have a chance to look even at validation report. In my opinion, whatever it is worth, every journal dealing with crystal structures should, at the very least, request the validation report from PDB including Nature, Science and PNAS. What is also interesting that at the end PDB released the coordinates with large number of outliers. I don’t think those can be justified with low resolution of the data. Regards, Vaheh Oganesyan www.medimmune.comUrlBlockedError.aspx From: CCP4 bulletin board [mailto:CCP4BB@JISCMAIL.AC.UK] On Behalf Of Robbie Joosten Sent: Friday, April 24, 2015 2:37 AM To: CCP4BB@JISCMAIL.AC.UK Subject: Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers! The PDB_REDO entry for 3bdn was pretty old, so I replaced it using a newer version of PDB_REDO that can use nucleic acid restraints from LibG: http://www.cmbi.ru.nl/pdb_redo/bd/3bdn/index.html. Obviously, the new structure model is far from brilliant (PDB_REDO doesn't rebuild at this resolution), but Molprobity seems to like it quite a bit more. I agree with the replies so far in that: - The topic starter was rather blunt and could have been more subtle. He should probably go work in the Netherlands ;) - Building structure models at 3.9A is incredibly difficult. - The tools we have now are much better than in 2008. However, we should not act like 2008 were still in the dark ages of crystallography. There are a lot of good structures available from that time (and also from long before) even at that resolution. That is not surprising seeing that we also already had very good building and refinement tools available. We also had enough validation tools available to tell us that this particular structure model isn't very good. I really believe that a good crystallographer that was not pressed for time (or at least didn't rush) could have done better with the data and the tools available. I'm now going to hide behind an asbestos wall to say this: The manuscript was submitted in July 29th 2007, the PDB entry was deposited November 15th 2007. That means that the referees probably did not have a chance to see the finished structure model, at least not in the first pass. This implies that the authors didn't want to deposit the model on time. There are a whole lot of excuses for this, that are fortunately dealt with now (http://onlinelibrary.wiley.com/doi/10.1107/S0907444913029168/abstract), but the referees could have been a bit more critical. They should have at least seen that the supplemental table 1 did not show any Ramachandran statistics. We can only speculate what happened. I'm guessing that the authors didn't finish the structure yet and rushed the publication through to avoid being scooped or for the general glory of a Nature paper. To bad that came at the expense of the crystallography. Cheers, Robbie Date: Thu, 23 Apr 2015 18:43:13 + From: kell...@janelia.hhmi.orgmailto:kell...@janelia.hhmi.org Subject: Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers! To: CCP4BB@JISCMAIL.AC.UKmailto:CCP4BB@JISCMAIL.AC.UK Is it in pdb redo? Take a look here: http://www.cmbi.ru.nl/pdb_redo/bd/3bdn/index.html JPK From: CCP4 bulletin board [mailto:CCP4BB@JISCMAIL.AC.UK] On Behalf Of Misbah ud Din Ahmad Sent: Thursday, April 23, 2015 2:28 PM To: CCP4BB@JISCMAIL.AC.UKmailto:CCP4BB@JISCMAIL.AC.UK Subject: Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers! Dear Phoebe A. Rice, I didn't mean to discredit the work but the statistics of the structure just shocked me at the first instance. I could for example point out to another structure 1ZR2, which has the same resolution (the protein Molecular weight is almost the same) and the statistics are: R-work: 0.27 Rfree: 0.319 Ramachandran outliers: 2.8% The structure was solved by a combination of MIR and MAD three years earlier in 2005 and definitely they didn't had better softwares. As per your advice, I tried one cycle of refinement of this structure in Phenix and the statistics are: Start R-work = 0.2816 ; Start R-free = 0.3551 Final R-work = 0.2887 ; Final R-free = 0.3671 Ramachandran outliers: 18.91% Rotamer Outliers: 25.19% Being a crystallographic community, isn't it our responsibility that if better softwares are available now, we try to re-refine our older structures and deposit better models in the PDB. That would help the users immensely. Misbha On Thu, Apr 23, 2015 at 7:06 PM, Mark J van Raaij
Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers!!!!!
Hi, Such kind of discussion are really great and fruitful. thanks Pankaj On Mon, Apr 27, 2015 at 12:25 PM, Dale Tronrud de...@daletronrud.com wrote: -BEGIN PGP SIGNED MESSAGE- Hash: SHA1 This particular model was deposited in early December of 2014, so the authors had the validation report in hand before publication and, I presume could (should) have passed it on to the reviewers. The release date of the entry has nothing to do with the availability of the validation report to reviewers. The model's validation report does have low percentiles on the clashscore and RSRZ outliers but is far better than most 3.5A models for Ramachandran outliers. I'm not sure what your objection to this model is. Did you look at the fit to density and have some specific criticism of the quality of workmanship? As for the PDB releasing a model containing outliers: The PDB is an archive, not a gatekeeper. If you, and your publisher, want to put your name on it the PDB will store it so that others can look at the model/data and judge for themselves. Dale Tronrud On 4/27/2015 7:04 AM, Oganesyan, Vaheh wrote: Hi Robbie and Co, These things are happening now too. Look at the entry 4x4m. The paper got published in January, PDB released coordinates in April. That means reviewers did not have a chance to look even at validation report. In my opinion, whatever it is worth, every journal dealing with crystal structures should, at the very least, request the validation report from PDB including Nature, Science and PNAS. What is also interesting that at the end PDB released the coordinates with large number of outliers. I don’t think those can be justified with low resolution of the data. /Regards,/ / / /Vaheh Oganesyan/ /www.medimmune.com/ *From:*CCP4 bulletin board [mailto:CCP4BB@JISCMAIL.AC.UK] *On Behalf Of *Robbie Joosten *Sent:* Friday, April 24, 2015 2:37 AM *To:* CCP4BB@JISCMAIL.AC.UK *Subject:* Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers! The PDB_REDO entry for 3bdn was pretty old, so I replaced it using a newer version of PDB_REDO that can use nucleic acid restraints from LibG: http://www.cmbi.ru.nl/pdb_redo/bd/3bdn/index.html. Obviously, the new structure model is far from brilliant (PDB_REDO doesn't rebuild at this resolution), but Molprobity seems to like it quite a bit more. I agree with the replies so far in that: - The topic starter was rather blunt and could have been more subtle. He should probably go work in the Netherlands ;) - Building structure models at 3.9A is incredibly difficult. - The tools we have now are much better than in 2008. However, we should not act like 2008 were still in the dark ages of crystallography. There are a lot of good structures available from that time (and also from long before) even at that resolution. That is not surprising seeing that we also already had very good building and refinement tools available. We also had enough validation tools available to tell us that this particular structure model isn't very good. I really believe that a good crystallographer that was not pressed for time (or at least didn't rush) could have done better with the data and the tools available. I'm now going to hide behind an asbestos wall to say this: The manuscript was submitted in July 29th 2007, the PDB entry was deposited November 15th 2007. That means that the referees probably did not have a chance to see the finished structure model, at least not in the first pass. This implies that the authors didn't want to deposit the model on time. There are a whole lot of excuses for this, that are fortunately dealt with now (http://onlinelibrary.wiley.com/doi/10.1107/S0907444913029168/abstract), but the referees could have been a bit more critical. They should have at least seen that the supplemental table 1 did not show any Ramachandran statistics. We can only speculate what happened. I'm guessing that the authors didn't finish the structure yet and rushed the publication through to avoid being scooped or for the general glory of a Nature paper. To bad that came at the expense of the crystallography. Cheers, Robbie Date: Thu, 23 Apr 2015 18:43:13 + From: kell...@janelia.hhmi.org mailto:kell...@janelia.hhmi.org Subject: Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers! To: CCP4BB@JISCMAIL.AC.UK mailto:CCP4BB@JISCMAIL.AC.UK Is it in pdb redo? Take a look here: http://www.cmbi.ru.nl/pdb_redo/bd/3bdn/index.html JPK *From:*CCP4 bulletin board [mailto:CCP4BB@JISCMAIL.AC.UK] *On Behalf Of *Misbah ud Din Ahmad *Sent:* Thursday, April 23, 2015 2:28 PM *To:* CCP4BB@JISCMAIL.AC.UK mailto:CCP4BB@JISCMAIL.AC.UK *Subject:* Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers! Dear Phoebe
Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers!!!!!
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 This particular model was deposited in early December of 2014, so the authors had the validation report in hand before publication and, I presume could (should) have passed it on to the reviewers. The release date of the entry has nothing to do with the availability of the validation report to reviewers. The model's validation report does have low percentiles on the clashscore and RSRZ outliers but is far better than most 3.5A models for Ramachandran outliers. I'm not sure what your objection to this model is. Did you look at the fit to density and have some specific criticism of the quality of workmanship? As for the PDB releasing a model containing outliers: The PDB is an archive, not a gatekeeper. If you, and your publisher, want to put your name on it the PDB will store it so that others can look at the model/data and judge for themselves. Dale Tronrud On 4/27/2015 7:04 AM, Oganesyan, Vaheh wrote: Hi Robbie and Co, These things are happening now too. Look at the entry 4x4m. The paper got published in January, PDB released coordinates in April. That means reviewers did not have a chance to look even at validation report. In my opinion, whatever it is worth, every journal dealing with crystal structures should, at the very least, request the validation report from PDB including Nature, Science and PNAS. What is also interesting that at the end PDB released the coordinates with large number of outliers. I don’t think those can be justified with low resolution of the data. /Regards,/ / / /Vaheh Oganesyan/ /www.medimmune.com/ *From:*CCP4 bulletin board [mailto:CCP4BB@JISCMAIL.AC.UK] *On Behalf Of *Robbie Joosten *Sent:* Friday, April 24, 2015 2:37 AM *To:* CCP4BB@JISCMAIL.AC.UK *Subject:* Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers! The PDB_REDO entry for 3bdn was pretty old, so I replaced it using a newer version of PDB_REDO that can use nucleic acid restraints from LibG: http://www.cmbi.ru.nl/pdb_redo/bd/3bdn/index.html. Obviously, the new structure model is far from brilliant (PDB_REDO doesn't rebuild at this resolution), but Molprobity seems to like it quite a bit more. I agree with the replies so far in that: - The topic starter was rather blunt and could have been more subtle. He should probably go work in the Netherlands ;) - Building structure models at 3.9A is incredibly difficult. - The tools we have now are much better than in 2008. However, we should not act like 2008 were still in the dark ages of crystallography. There are a lot of good structures available from that time (and also from long before) even at that resolution. That is not surprising seeing that we also already had very good building and refinement tools available. We also had enough validation tools available to tell us that this particular structure model isn't very good. I really believe that a good crystallographer that was not pressed for time (or at least didn't rush) could have done better with the data and the tools available. I'm now going to hide behind an asbestos wall to say this: The manuscript was submitted in July 29th 2007, the PDB entry was deposited November 15th 2007. That means that the referees probably did not have a chance to see the finished structure model, at least not in the first pass. This implies that the authors didn't want to deposit the model on time. There are a whole lot of excuses for this, that are fortunately dealt with now (http://onlinelibrary.wiley.com/doi/10.1107/S0907444913029168/abstract), but the referees could have been a bit more critical. They should have at least seen that the supplemental table 1 did not show any Ramachandran statistics. We can only speculate what happened. I'm guessing that the authors didn't finish the structure yet and rushed the publication through to avoid being scooped or for the general glory of a Nature paper. To bad that came at the expense of the crystallography. Cheers, Robbie Date: Thu, 23 Apr 2015 18:43:13 + From: kell...@janelia.hhmi.org mailto:kell...@janelia.hhmi.org Subject: Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers! To: CCP4BB@JISCMAIL.AC.UK mailto:CCP4BB@JISCMAIL.AC.UK Is it in pdb redo? Take a look here: http://www.cmbi.ru.nl/pdb_redo/bd/3bdn/index.html JPK *From:*CCP4 bulletin board [mailto:CCP4BB@JISCMAIL.AC.UK] *On Behalf Of *Misbah ud Din Ahmad *Sent:* Thursday, April 23, 2015 2:28 PM *To:* CCP4BB@JISCMAIL.AC.UK mailto:CCP4BB@JISCMAIL.AC.UK *Subject:* Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers! Dear Phoebe A. Rice, I didn't mean to discredit the work but the statistics of the structure just shocked me at the first instance. I could for example point out to another structure 1ZR2, which has the same resolution (the protein Molecular
Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers!!!!!
Hi Robbie: Sorry, I just saw your email now. To make that figure, I just downloaded the automated refinement results manually from the server, loaded into coot, and took a screenshot. A couple of days later, after Gert mentioned you had since fixed it, I checked with the built-in coot interface, and, as you say, it looks fine now. In general, the problem seems to do with linking. Two examples (which both now appear to be corrected) in 3zp8 are the 5’-terminal GDP (one often gets GTP or GDP at the 5’-end of in vitro transcribed RNAs) and the modified nucleotide at the ribozyme cleavage site (2’-OMe). In some instances, these don’t get linked up the the rest of the nucleotide chain. I think the root problem is different refinement programs handle some of these things more or less gracefully by default, probably because the PDB standard (if there is one) is obscure. For example, GDP (or GTP) is often a separate cofactor, so I guess it doesn’t get linked by default. I’ve run into different subsets of this problem both with refmac and with phenix, although refmac run via coot seems to handle it gracefully. Phenix complains it can’t find the 2’-OMC cif file by default, but when pointed in the right direction, handles it, but it needs to be told to link the GDP at the 5’-terminus. PDB_REDO is a great service; certainly there is no reason to apologize for the results of an automated refinement. If it “goes wrong”, this really tells us that there is a more fundamental problem of standardization that needs to be addressed, because the automatic refinment program should run without any special knowledge, just as any interested user should be able to take the PDB and data and refine the structure for himself or herself manually without any special knowledge. All the best, Bill William G. Scott Professor, Department of Chemistry and Biochemistry and The Center for the Molecular Biology of RNA University of California at Santa Cruz Santa Cruz, California 95064 USA http://scottlab.ucsc.edu/~wgscott On Apr 24, 2015, at 12:52 AM, Robbie Joosten robbie_joos...@hotmail.com wrote: Hi Bill, I have no Idea where you got that picture, the PDB_REDO entry for 3zp8 look pretty okay to me. You might need press F5 (or the OSX equivalent) when you check the entry page though ;) Apparently something went hideously wrong with the restraint generation in the previous run. It seems to work much better now. The modified residue stays nicely in density. Problems like this are bound to occur in automated pipelines like PDB_REDO, we cannot check all the entries by hand. So if you or anyone else finds problems with specific entries, feedback is always welcomed. We might be able to solve the problem and improve out software in the process. Cheers, Robbie Date: Thu, 23 Apr 2015 16:15:41 -0700 From: wgsc...@ucsc.edu Subject: Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers! To: CCP4BB@JISCMAIL.AC.UK On Apr 23, 2015, at 11:43 AM, Keller, Jacob kell...@janelia.hhmi.org wrote: Is it in pdb redo? Take a look here: http://www.cmbi.ru.nl/pdb_redo/bd/3bdn/index.html JPK Sometimes it is worth checking the pdb_redo maps, in case it does something like it did to our 3zp8 (which has a modified nucleotide at the active site that apparently it can’t cope with): Screen Shot 2015-04-23 at 4.02.46 PM.png Maybe it just hates RNA and defaults to pdb_oyvey Bill William G. Scott http://scottlab.ucsc.edu/~wgscott
Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers!!!!!
Ramachandran outliers aside, amount of other outliers seem a bit worrying to me, especially given that it may not be all that trivial to justify them against 3.5A resolution map: all-atom clashscore : 29.19 ramachandran plot: outliers : 0.51 % allowed : 5.44 % favored : 94.05 % rotamer outliers : 8.23 % cbeta deviations : 5 (Command used to get the numbers above: phenix.pdbtools model_stat=true 4x4m.pdb ) Pavel On Mon, Apr 27, 2015 at 7:54 AM, Todd Jason Green tgr...@uab.edu wrote: 7 out of 1361 Ramachandran outliers (all prolines) doesn't seem high to me. -- *From:* CCP4 bulletin board [CCP4BB@JISCMAIL.AC.UK] on behalf of Oganesyan, Vaheh [oganesy...@medimmune.com] *Sent:* Monday, April 27, 2015 9:04 AM *To:* CCP4BB@JISCMAIL.AC.UK *Subject:* Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers! Hi Robbie and Co, These things are happening now too. Look at the entry 4x4m. The paper got published in January, PDB released coordinates in April. That means reviewers did not have a chance to look even at validation report. In my opinion, whatever it is worth, every journal dealing with crystal structures should, at the very least, request the validation report from PDB including Nature, Science and PNAS. What is also interesting that at the end PDB released the coordinates with large number of outliers. I don’t think those can be justified with low resolution of the data.
Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers!!!!!
Sorry can't help it. The aggressive replies are mainly from senior PIs from US - friend in need is friend indeed. On Thu, Apr 23, 2015 at 2:20 PM, Bernhard Rupp (Hofkristallrat a.D.) hofkristall...@gmail.com wrote: Well, with full respect to your sensitivities as far as your own person is concerned: if you play rough (and 5 exclamation marks qualify by commonly accepted email etiquette as incipient flame, at least), you need to be willing to take a few as well... Best, BR - Trigger warning --- This message may or may not contain references to issues of privilege and oppression in- cluding but not limited to enantiophobia, point classism, heteroatomism, transbondism, cisbonding, sizeism, curvature, references to color, alcohol, blood, small insects, cancer cells, and any combination thereof that may cause symptoms ranging from discomfort and anxiety to violent physical response in sensitive individuals and must therefore be labeled as potentially hazardous to your comfort and well-being. - Trigger warning --- *From:* CCP4 bulletin board [mailto:CCP4BB@JISCMAIL.AC.UK] *On Behalf Of *Misbah ud Din Ahmad *Sent:* Thursday, April 23, 2015 2:05 PM *To:* CCP4BB@JISCMAIL.AC.UK *Subject:* Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers! I feel rather uncomfortable and cornered with the aggressive replies and phrases like throwing stones, public humiliation etc., which suggest that I have some personal enmity with the author. The structure is in the public domain and questions can be asked about it. I as a beginner in crystallography learn from each discussion on this board and such harsh and insinuating responses may, in future, deter people like me from posting any questions at all and learning refinement strategies like the one which Pavel outlined to improve such type of structures. Best Misbha On Thu, Apr 23, 2015 at 9:17 PM, Gert Vriend gerrit.vri...@radboudumc.nl wrote: At around 4.0 A resolution one normally cannot talk about accuracy. The density will at most locations not warrant any detailed interpretation. If, at 4.0 A resolution, you move atoms around a bit, you will not see significant changes in R/Rfree. So, you can do whatever you want, more or less. If the refinement puts great emphasis on the secondary structure (i.e. tries to force the Ramachandran plot, then you get a good Ramachandran plot, but if they put more weight of the fit to the density, you get a slightly lower R (and perhaps even Rfree) at the cost of the Ramachandran plot. And all of that is meaningless. At this resolution you have to check if the fold is likely to be correct. I superposed just any high(er) resolution domain with high sequence similarity at the corresponding 3bdn domain, and I take any bet that the fold of 3bdn is right (at least of the domain I looked at, and I extrapolate to the other domain). PDB_REDO (http://www.cmbi.ru.nl/pdb_redo/) cannot improve this one, and WHAT_CHECK (http://swift.cmbi.ru.nl/gv/pdbreport/) throws its hands up in the air. At 4.0 A you should be happy with an indication of the fold (and with the fact that the authors probably went through great pains for you getting these coordinates close to where they should be). Gert
Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers!!!!!
The PDB_REDO entry for 3bdn was pretty old, so I replaced it using a newer version of PDB_REDO that can use nucleic acid restraints from LibG: http://www.cmbi.ru.nl/pdb_redo/bd/3bdn/index.html. Obviously, the new structure model is far from brilliant (PDB_REDO doesn't rebuild at this resolution), but Molprobity seems to like it quite a bit more. I agree with the replies so far in that: - The topic starter was rather blunt and could have been more subtle. He should probably go work in the Netherlands ;) - Building structure models at 3.9A is incredibly difficult. - The tools we have now are much better than in 2008. However, we should not act like 2008 were still in the dark ages of crystallography. There are a lot of good structures available from that time (and also from long before) even at that resolution. That is not surprising seeing that we also already had very good building and refinement tools available. We also had enough validation tools available to tell us that this particular structure model isn't very good. I really believe that a good crystallographer that was not pressed for time (or at least didn't rush) could have done better with the data and the tools available. I'm now going to hide behind an asbestos wall to say this: The manuscript was submitted in July 29th 2007, the PDB entry was deposited November 15th 2007. That means that the referees probably did not have a chance to see the finished structure model, at least not in the first pass. This implies that the authors didn't want to deposit the model on time. There are a whole lot of excuses for this, that are fortunately dealt with now (http://onlinelibrary.wiley.com/doi/10.1107/S0907444913029168/abstract), but the referees could have been a bit more critical. They should have at least seen that the supplemental table 1 did not show any Ramachandran statistics. We can only speculate what happened. I'm guessing that the authors didn't finish the structure yet and rushed the publication through to avoid being scooped or for the general glory of a Nature paper. To bad that came at the expense of the crystallography. Cheers, Robbie Date: Thu, 23 Apr 2015 18:43:13 + From: kell...@janelia.hhmi.org Subject: Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers! To: CCP4BB@JISCMAIL.AC.UK Is it in pdb redo? Take a look here: http://www.cmbi.ru.nl/pdb_redo/bd/3bdn/index.html JPK From: CCP4 bulletin board [mailto:CCP4BB@JISCMAIL.AC.UK] On Behalf Of Misbah ud Din Ahmad Sent: Thursday, April 23, 2015 2:28 PM To: CCP4BB@JISCMAIL.AC.UK Subject: Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers! Dear Phoebe A. Rice, I didn't mean to discredit the work but the statistics of the structure just shocked me at the first instance. I could for example point out to another structure 1ZR2, which has the same resolution (the protein Molecular weight is almost the same) and the statistics are: R-work: 0.27 Rfree: 0.319 Ramachandran outliers: 2.8% The structure was solved by a combination of MIR and MAD three years earlier in 2005 and definitely they didn't had better softwares. As per your advice, I tried one cycle of refinement of this structure in Phenix and the statistics are: Start R-work = 0.2816 ; Start R-free = 0.3551 Final R-work = 0.2887 ; Final R-free = 0.3671 Ramachandran outliers: 18.91% Rotamer Outliers: 25.19% Being a crystallographic community, isn't it our responsibility that if better softwares are available now, we try to re-refine our older structures and deposit better models in the PDB. That would help the users immensely. Misbha On Thu, Apr 23, 2015 at 7:06 PM, Mark J van Raaij mjvanra...@cnb.csic.es wrote: The abstract of the papers says they used MIR. Mark J van Raaij Dpto de Estructura de Macromoleculas Centro Nacional de Biotecnologia - CSIC c/Darwin 3 E-28049 Madrid, Spain tel. (+34) 91 585 4616 http://www.cnb.csic.es/~mjvanraaij On 23 Apr 2015, at 18:57, Todd Jason Green wrote: My guess is they had the best data they could get, did molecular replacement with the two halves of the repressor and the dna, got a solution and didn't use appropriate restraints in the refinement. Like Phoebe mentioned, we have better tools for this these days. From: CCP4 bulletin board [CCP4BB@JISCMAIL.AC.UK] on behalf of Mark J van Raaij [mjvanra...@cnb.csic.es] Sent: Thursday, April 23, 2015 11:49 AM To: CCP4BB@JISCMAIL.AC.UK Subject: Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers! How reliable is too general a question - it depends on what you want to know. At 3.9Å they could probably place the phosphate atoms quite well and see the general fold of the protein. Finer details will be less reliable, i.e. where the exact side-chains are etc. They could probably have forced more amino acids into favourable
Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers!!!!!
Dear Mishba, Just check density vs model by simply open 'Coot', Go to: 'File' - 'Fetch PDB Map using EDS' Type the pdb entry into the field - enjoy the densities. Best and god save the EDS, Gert - LOEWE Center for Synthetic Microbiology AG Bange - Analysis of Metabolic Networks Dr. Gert Bange Hans-Meerwein-Strasse, C7 35043 Marburg, Germany office: +49-6421-28-23361 fax: +49-6421-28-24430 web: www.synmikro.com/bange - LOEWE Center for Synthetic Microbiology AG Bange - Analysis of Metabolic Networks Dr. Gert Bange Hans-Meerwein-Strasse, C7 35043 Marburg, Germany office: +49-6421-28-23361 fax: +49-6421-28-24430 web: www.synmikro.com/bange Am 23.04.2015 um 18:03 schrieb Misbah ud Din Ahmad misba.ah...@gmail.com: Dear crystallographers, The PDB entry http://www.rcsb.org/pdb/explore.do?structureId=3BDN has 16.5% Ramachandran outliers. When I opened this PDB file in coot and checked for Ramachandran outliers, the results are: In preffered region: 58.04% In allowed regions: 19.78% Outliers: 22.17% ! With an R-free of 37.4% at 3.9 A resolution, could you please tell me how reliable this structure of Lambda repressor bound to DNA is? Thanks Misbha
Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers!!!!!
It was published in Nature so it must be right :) ... Dear crystallographers, The PDB entry http://www.rcsb.org/pdb/explore.do?structureId=3BDNhttp://www.rcsb.org/pdb/explore.do?structureId=3BDN has 16.5% Ramachandran outliers. When I opened this PDB file in coot and checked for Ramachandran outliers, the results are: In preffered region: 58.04% In allowed regions: 19.78% Outliers: 22.17% ! With an R-free of 37.4% at 3.9 A resolution, could you please tell me how reliable this structure of Lambda repressor bound to DNA is? Thanks Misbha -- +++ Mark Mayer Ph.D. LCMN NICHD NIH DHHS Bldg 35, Room 3B 1002 MSC 3712 35 Lincoln Drive Bethesda MD 20892 3712 Phone: 301-496-9346 (office); 301-496-9347 (lab); FAX 301-496-2396 Lab web site: http://snb.nichd.nih.gov/index.htm Send packages, Fedex and anything requiring a signature to: Bldg 35, Room 3B 1004 35 Lincoln Drive Bethesda MD 20892
Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers!!!!!
Dear Gert, I just did that and I wonder how these structures end there !!! @ Sabine: I was talking about Ramachandran outliers. I had learnt that the number of Ramachandran outliers show be very few or none for low resolution structures. On Thu, Apr 23, 2015 at 6:25 PM, Gert Bange gert.ba...@synmikro.uni-marburg.de wrote: Dear Mishba, Just check density vs model by simply open 'Coot', Go to: 'File' - 'Fetch PDB Map using EDS' Type the pdb entry into the field - enjoy the densities. Best and god save the EDS, Gert - LOEWE Center for Synthetic Microbiology AG Bange - Analysis of Metabolic Networks Dr. Gert Bange Hans-Meerwein-Strasse, C7 35043 Marburg, Germany office: +49-6421-28-23361 fax: +49-6421-28-24430 web: www.synmikro.com/bange - LOEWE Center for Synthetic Microbiology AG Bange - Analysis of Metabolic Networks Dr. Gert Bange Hans-Meerwein-Strasse, C7 35043 Marburg, Germany office: +49-6421-28-23361 fax: +49-6421-28-24430 web: www.synmikro.com/bange Am 23.04.2015 um 18:03 schrieb Misbah ud Din Ahmad misba.ah...@gmail.com : Dear crystallographers, The PDB entry http://www.rcsb.org/pdb/explore.do?structureId=3BDN has 16.5% Ramachandran outliers. When I opened this PDB file in coot and checked for Ramachandran outliers, the results are: In preffered region: 58.04% In allowed regions: 19.78% *Outliers: 22.17% !* With an R-free of 37.4% at 3.9 A resolution, could you please tell me how reliable this structure of Lambda repressor bound to DNA is? Thanks Misbha
Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers!!!!!
They have, it is call PDB_REDO. You can download the maps and refined structure there. http://www.cmbi.ru.nl/pdb_redo/ Dan
Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers!!!!!
Its not just the Ramachandran outliers - the attached metrics summary courtesy of the PDB raises a lot of questions about the accuracy of this structure As for Nature being right, it was an attempt at humor. Dear crystallographers, The PDB entry http://www.rcsb.org/pdb/explore.do?structureId=3BDNhttp://www.rcsb.org/pdb/explore.do?structureId=3BDN has 16.5% Ramachandran outliers. When I opened this PDB file in coot and checked for Ramachandran outliers, the results are: In preffered region: 58.04% In allowed regions: 19.78% Outliers: 22.17% ! With an R-free of 37.4% at 3.9 A resolution, could you please tell me how reliable this structure of Lambda repressor bound to DNA is? Thanks Misbha -- +++ Mark Mayer Ph.D. LCMN NICHD NIH DHHS Bldg 35, Room 3B 1002 MSC 3712 35 Lincoln Drive Bethesda MD 20892 3712 Phone: 301-496-9346 (office); 301-496-9347 (lab); FAX 301-496-2396 Lab web site: http://snb.nichd.nih.gov/index.htm Send packages, Fedex and anything requiring a signature to: Bldg 35, Room 3B 1004 35 Lincoln Drive Bethesda MD 20892 3BDN_metrics.png Description: Binary data
Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers!!!!!
Is it in pdb redo? Take a look here: http://www.cmbi.ru.nl/pdb_redo/bd/3bdn/index.html JPK From: CCP4 bulletin board [mailto:CCP4BB@JISCMAIL.AC.UK] On Behalf Of Misbah ud Din Ahmad Sent: Thursday, April 23, 2015 2:28 PM To: CCP4BB@JISCMAIL.AC.UK Subject: Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers! Dear Phoebe A. Rice, I didn't mean to discredit the work but the statistics of the structure just shocked me at the first instance. I could for example point out to another structure 1ZR2, which has the same resolution (the protein Molecular weight is almost the same) and the statistics are: R-work: 0.27 Rfree: 0.319 Ramachandran outliers: 2.8% The structure was solved by a combination of MIR and MAD three years earlier in 2005 and definitely they didn't had better softwares. As per your advice, I tried one cycle of refinement of this structure in Phenix and the statistics are: Start R-work = 0.2816 ; Start R-free = 0.3551 Final R-work = 0.2887 ; Final R-free = 0.3671 Ramachandran outliers: 18.91% Rotamer Outliers: 25.19% Being a crystallographic community, isn't it our responsibility that if better softwares are available now, we try to re-refine our older structures and deposit better models in the PDB. That would help the users immensely. Misbha On Thu, Apr 23, 2015 at 7:06 PM, Mark J van Raaij mjvanra...@cnb.csic.esmailto:mjvanra...@cnb.csic.es wrote: The abstract of the papers says they used MIR. Mark J van Raaij Dpto de Estructura de Macromoleculas Centro Nacional de Biotecnologia - CSIC c/Darwin 3 E-28049 Madrid, Spain tel. (+34) 91 585 4616tel:%28%2B34%29%2091%20585%204616 http://www.cnb.csic.es/~mjvanraaij On 23 Apr 2015, at 18:57, Todd Jason Green wrote: My guess is they had the best data they could get, did molecular replacement with the two halves of the repressor and the dna, got a solution and didn't use appropriate restraints in the refinement. Like Phoebe mentioned, we have better tools for this these days. From: CCP4 bulletin board [CCP4BB@JISCMAIL.AC.UKmailto:CCP4BB@JISCMAIL.AC.UK] on behalf of Mark J van Raaij [mjvanra...@cnb.csic.esmailto:mjvanra...@cnb.csic.es] Sent: Thursday, April 23, 2015 11:49 AM To: CCP4BB@JISCMAIL.AC.UKmailto:CCP4BB@JISCMAIL.AC.UK Subject: Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers! How reliable is too general a question - it depends on what you want to know. At 3.9Å they could probably place the phosphate atoms quite well and see the general fold of the protein. Finer details will be less reliable, i.e. where the exact side-chains are etc. They could probably have forced more amino acids into favourable Ramachandran angles, but would that have made the structure better? Would these favourable angles have been more right? At 3.9Å you can't know for sure. Would they have been able to draw more biological conclusions? I'd say not. As long as they do not draw more conclusions in the paper than what is supported by the medium-resolution data, the structure provides useful information. Mark J van Raaij Dpto de Estructura de Macromoleculas Centro Nacional de Biotecnologia - CSIC c/Darwin 3 E-28049 Madrid, Spain tel. (+34) 91 585 4616 http://www.cnb.csic.es/~mjvanraaij On 23 Apr 2015, at 18:03, Misbah ud Din Ahmad wrote: Dear crystallographers, The PDB entry http://www.rcsb.org/pdb/explore.do?structureId=3BDN has 16.5% Ramachandran outliers. When I opened this PDB file in coot and checked for Ramachandran outliers, the results are: In preffered region: 58.04% In allowed regions: 19.78% Outliers: 22.17% ! With an R-free of 37.4% at 3.9 A resolution, could you please tell me how reliable this structure of Lambda repressor bound to DNA is? Thanks Misbha
Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers!!!!!
My guess is they had the best data they could get, did molecular replacement with the two halves of the repressor and the dna, got a solution and didn't use appropriate restraints in the refinement. Like Phoebe mentioned, we have better tools for this these days. From: CCP4 bulletin board [CCP4BB@JISCMAIL.AC.UK] on behalf of Mark J van Raaij [mjvanra...@cnb.csic.es] Sent: Thursday, April 23, 2015 11:49 AM To: CCP4BB@JISCMAIL.AC.UK Subject: Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers! How reliable is too general a question - it depends on what you want to know. At 3.9Å they could probably place the phosphate atoms quite well and see the general fold of the protein. Finer details will be less reliable, i.e. where the exact side-chains are etc. They could probably have forced more amino acids into favourable Ramachandran angles, but would that have made the structure better? Would these favourable angles have been more right? At 3.9Å you can't know for sure. Would they have been able to draw more biological conclusions? I'd say not. As long as they do not draw more conclusions in the paper than what is supported by the medium-resolution data, the structure provides useful information. Mark J van Raaij Dpto de Estructura de Macromoleculas Centro Nacional de Biotecnologia - CSIC c/Darwin 3 E-28049 Madrid, Spain tel. (+34) 91 585 4616 http://www.cnb.csic.es/~mjvanraaij On 23 Apr 2015, at 18:03, Misbah ud Din Ahmad wrote: Dear crystallographers, The PDB entry http://www.rcsb.org/pdb/explore.do?structureId=3BDN has 16.5% Ramachandran outliers. When I opened this PDB file in coot and checked for Ramachandran outliers, the results are: In preffered region: 58.04% In allowed regions: 19.78% Outliers: 22.17% ! With an R-free of 37.4% at 3.9 A resolution, could you please tell me how reliable this structure of Lambda repressor bound to DNA is? Thanks Misbha
Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers!!!!!
Dear Misbha, Before you throw stones at other peoples' work: - have YOU tried to refine a structure at 3.9A, particularly with older software? - there are many structures in the database that are related to lambda repressor. Have you checked how the structure in question compares to those determined at higher resolution? Is the overall fold the same? Is the sequence register the same? Remember that it is unreasonable to expect that all the side chain rotamers will be correct at 3.9A. - When you looked at the actual electron density, what, exactly, did you see that suggested the structure may not be correct? - The structure factors were deposited, so it should be easy for you to try re-refining it yourself with newer software, and see if it makes much difference to the final model. - If you've gone through all those steps, and are still convinced that this structure is problematic, please do let the community know! ++ Phoebe A. Rice Dept. of Biochemistry Molecular Biology The University of Chicago pr...@uchicago.edumailto:pr...@uchicago.edu From: CCP4 bulletin board [CCP4BB@JISCMAIL.AC.UK] on behalf of Misbah ud Din Ahmad [misba.ah...@gmail.com] Sent: Thursday, April 23, 2015 11:03 AM To: CCP4BB@JISCMAIL.AC.UK Subject: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers! Dear crystallographers, The PDB entry http://www.rcsb.org/pdb/explore.do?structureId=3BDN has 16.5% Ramachandran outliers. When I opened this PDB file in coot and checked for Ramachandran outliers, the results are: In preffered region: 58.04% In allowed regions: 19.78% Outliers: 22.17% ! With an R-free of 37.4% at 3.9 A resolution, could you please tell me how reliable this structure of Lambda repressor bound to DNA is? Thanks Misbha
Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers!!!!!
On Thursday, 23 April, 2015 18:25:35 Gert Bange wrote: Dear Mishba, Just check density vs model by simply open 'Coot', Go to: 'File' - 'Fetch PDB Map using EDS' Type the pdb entry into the field - enjoy the densities. If you want to pursue the plausibility of specific outliers you could also download and superimpose PDB entry 1LMB (deposited in 1993, 15 years earlier) and calculate a Kleywegt plot comparing the difference in phi/psi between the two structures at each residue they share. This is in the validation menu of Coot. Compare the models at their points of difference and decide for yourself. 1LMB has 0% phi/psi outliers, though it contains a smaller piece of the structure. If you care to dig deeper, you could re-refine 3BDN after substituting the 1LMB coordinates for the 3BDN coordinates in the region where the chains superimpose well. I have not looked at any of the other related PDB structures, but you would want investigate those as well. Ethan Best and god save the EDS, Gert - LOEWE Center for Synthetic Microbiology AG Bange - Analysis of Metabolic Networks Dr. Gert Bange Hans-Meerwein-Strasse, C7 35043 Marburg, Germany office: +49-6421-28-23361 fax: +49-6421-28-24430 web: www.synmikro.com/bange - LOEWE Center for Synthetic Microbiology AG Bange - Analysis of Metabolic Networks Dr. Gert Bange Hans-Meerwein-Strasse, C7 35043 Marburg, Germany office: +49-6421-28-23361 fax: +49-6421-28-24430 web: www.synmikro.com/bange Am 23.04.2015 um 18:03 schrieb Misbah ud Din Ahmad misba.ah...@gmail.com: Dear crystallographers, The PDB entry http://www.rcsb.org/pdb/explore.do?structureId=3BDN has 16.5% Ramachandran outliers. When I opened this PDB file in coot and checked for Ramachandran outliers, the results are: In preffered region: 58.04% In allowed regions: 19.78% Outliers: 22.17% ! With an R-free of 37.4% at 3.9 A resolution, could you please tell me how reliable this structure of Lambda repressor bound to DNA is? Thanks Misbha -- Ethan A Merritt Biomolecular Structure Center, K-428 Health Sciences Bldg MS 357742, University of Washington, Seattle 98195-7742
Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers!!!!!
The methods part of the 2008 nature paper also mentioned the two earlier, high-res structures 1F39 (1.9 A) and 1LMB (1.8 A), which together cover almost the whole strucutre except the linker regions. I wonder, in such situations, is it a good practise to use the high res structures, either as referenece structures or as start points for refinement, to improve the low res structures? I have no doubt that most people would give both a try and compare the results. But what are people's opinions on publishing the resulting low res structures that contain significant amount of information coming from prior high-res strucutures? In which way should we think: this gives better structure so that the user sees less artifact or, no, this introduces information that did not come from my data? Zhijie -Original Message- From: Mark J van Raaij Sent: Thursday, April 23, 2015 1:05 PM To: CCP4BB@JISCMAIL.AC.UK Subject: Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers! The abstract of the papers says they used MIR. Mark J van Raaij Dpto de Estructura de Macromoleculas Centro Nacional de Biotecnologia - CSIC c/Darwin 3 E-28049 Madrid, Spain tel. (+34) 91 585 4616 http://www.cnb.csic.es/~mjvanraaij On 23 Apr 2015, at 18:57, Todd Jason Green wrote: My guess is they had the best data they could get, did molecular replacement with the two halves of the repressor and the dna, got a solution and didn't use appropriate restraints in the refinement. Like Phoebe mentioned, we have better tools for this these days. From: CCP4 bulletin board [CCP4BB@JISCMAIL.AC.UK] on behalf of Mark J van Raaij [mjvanra...@cnb.csic.es] Sent: Thursday, April 23, 2015 11:49 AM To: CCP4BB@JISCMAIL.AC.UK Subject: Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers! How reliable is too general a question - it depends on what you want to know. At 3.9Å they could probably place the phosphate atoms quite well and see the general fold of the protein. Finer details will be less reliable, i.e. where the exact side-chains are etc. They could probably have forced more amino acids into favourable Ramachandran angles, but would that have made the structure better? Would these favourable angles have been more right? At 3.9Å you can't know for sure. Would they have been able to draw more biological conclusions? I'd say not. As long as they do not draw more conclusions in the paper than what is supported by the medium-resolution data, the structure provides useful information. Mark J van Raaij Dpto de Estructura de Macromoleculas Centro Nacional de Biotecnologia - CSIC c/Darwin 3 E-28049 Madrid, Spain tel. (+34) 91 585 4616 http://www.cnb.csic.es/~mjvanraaij On 23 Apr 2015, at 18:03, Misbah ud Din Ahmad wrote: Dear crystallographers, The PDB entry http://www.rcsb.org/pdb/explore.do?structureId=3BDN has 16.5% Ramachandran outliers. When I opened this PDB file in coot and checked for Ramachandran outliers, the results are: In preffered region: 58.04% In allowed regions: 19.78% Outliers: 22.17% ! With an R-free of 37.4% at 3.9 A resolution, could you please tell me how reliable this structure of Lambda repressor bound to DNA is? Thanks Misbha
Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers!!!!!
Dear Phoebe A. Rice, I didn't mean to discredit the work but the statistics of the structure just shocked me at the first instance. I could for example point out to another structure 1ZR2, which has the same resolution (the protein Molecular weight is almost the same) and the statistics are: R-work: 0.27 Rfree: 0.319 Ramachandran outliers: 2.8% The structure was solved by a combination of MIR and MAD three years earlier in 2005 and definitely they didn't had better softwares. As per your advice, I tried one cycle of refinement of this structure in Phenix and the statistics are: Start R-work = 0.2816 ; Start R-free =* 0.3551* Final R-work = 0.2887 ; Final R-free = *0.3671* Ramachandran outliers: 18.91% Rotamer Outliers: 25.19% Being a crystallographic community, isn't it our responsibility that if better softwares are available now, we try to re-refine our older structures and deposit better models in the PDB. That would help the users immensely. Misbha On Thu, Apr 23, 2015 at 7:06 PM, Mark J van Raaij mjvanra...@cnb.csic.es wrote: The abstract of the papers says they used MIR. Mark J van Raaij Dpto de Estructura de Macromoleculas Centro Nacional de Biotecnologia - CSIC c/Darwin 3 E-28049 Madrid, Spain tel. (+34) 91 585 4616 http://www.cnb.csic.es/~mjvanraaij On 23 Apr 2015, at 18:57, Todd Jason Green wrote: My guess is they had the best data they could get, did molecular replacement with the two halves of the repressor and the dna, got a solution and didn't use appropriate restraints in the refinement. Like Phoebe mentioned, we have better tools for this these days. From: CCP4 bulletin board [CCP4BB@JISCMAIL.AC.UK] on behalf of Mark J van Raaij [mjvanra...@cnb.csic.es] Sent: Thursday, April 23, 2015 11:49 AM To: CCP4BB@JISCMAIL.AC.UK Subject: Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers! How reliable is too general a question - it depends on what you want to know. At 3.9Å they could probably place the phosphate atoms quite well and see the general fold of the protein. Finer details will be less reliable, i.e. where the exact side-chains are etc. They could probably have forced more amino acids into favourable Ramachandran angles, but would that have made the structure better? Would these favourable angles have been more right? At 3.9Å you can't know for sure. Would they have been able to draw more biological conclusions? I'd say not. As long as they do not draw more conclusions in the paper than what is supported by the medium-resolution data, the structure provides useful information. Mark J van Raaij Dpto de Estructura de Macromoleculas Centro Nacional de Biotecnologia - CSIC c/Darwin 3 E-28049 Madrid, Spain tel. (+34) 91 585 4616 http://www.cnb.csic.es/~mjvanraaij On 23 Apr 2015, at 18:03, Misbah ud Din Ahmad wrote: Dear crystallographers, The PDB entry http://www.rcsb.org/pdb/explore.do?structureId=3BDN has 16.5% Ramachandran outliers. When I opened this PDB file in coot and checked for Ramachandran outliers, the results are: In preffered region: 58.04% In allowed regions: 19.78% Outliers: 22.17% ! With an R-free of 37.4% at 3.9 A resolution, could you please tell me how reliable this structure of Lambda repressor bound to DNA is? Thanks Misbha
Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers!!!!!
Isn't the only criteria for EDS that calculated R (by EDS) is +/- 5% of reported R? On Thu, Apr 23, 2015 at 12:35 PM, Misbah ud Din Ahmad misba.ah...@gmail.com wrote: Dear Gert, I just did that and I wonder how these structures end there !!! @ Sabine: I was talking about Ramachandran outliers. I had learnt that the number of Ramachandran outliers show be very few or none for low resolution structures. On Thu, Apr 23, 2015 at 6:25 PM, Gert Bange gert.ba...@synmikro.uni-marburg.de wrote: Dear Mishba, Just check density vs model by simply open 'Coot', Go to: 'File' - 'Fetch PDB Map using EDS' Type the pdb entry into the field - enjoy the densities. Best and god save the EDS, Gert - LOEWE Center for Synthetic Microbiology AG Bange - Analysis of Metabolic Networks Dr. Gert Bange Hans-Meerwein-Strasse, C7 35043 Marburg, Germany office: +49-6421-28-23361 fax: +49-6421-28-24430 web: www.synmikro.com/bange - LOEWE Center for Synthetic Microbiology AG Bange - Analysis of Metabolic Networks Dr. Gert Bange Hans-Meerwein-Strasse, C7 35043 Marburg, Germany office: +49-6421-28-23361 fax: +49-6421-28-24430 web: www.synmikro.com/bange Am 23.04.2015 um 18:03 schrieb Misbah ud Din Ahmad misba.ah...@gmail.com : Dear crystallographers, The PDB entry http://www.rcsb.org/pdb/explore.do?structureId=3BDN has 16.5% Ramachandran outliers. When I opened this PDB file in coot and checked for Ramachandran outliers, the results are: In preffered region: 58.04% In allowed regions: 19.78% *Outliers: 22.17% !* With an R-free of 37.4% at 3.9 A resolution, could you please tell me how reliable this structure of Lambda repressor bound to DNA is? Thanks Misbha
Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers!!!!!
Elaborating on Phoebe's answer, the question to ask is : is the structure good enough to support the conclusions reached in the paper ? That is, quoting the abstract : an unusual overall architecture that allows it to adopt a conformation that appears to facilitate pairwise cooperative binding to adjacent operator sites. The answer is yes. Another question could be: does the structure provides an accurate picture of the structure, within a 0.2Å positional error. The answer is no. But then the authors made no such claim. Thierry From: CCP4 bulletin board [mailto:CCP4BB@JISCMAIL.AC.UK] On Behalf Of Phoebe A. Rice Sent: Thursday, April 23, 2015 12:46 PM To: CCP4BB@JISCMAIL.AC.UK Subject: Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers! Dear Misbha, Before you throw stones at other peoples' work: - have YOU tried to refine a structure at 3.9A, particularly with older software? - there are many structures in the database that are related to lambda repressor. Have you checked how the structure in question compares to those determined at higher resolution? Is the overall fold the same? Is the sequence register the same? Remember that it is unreasonable to expect that all the side chain rotamers will be correct at 3.9A. - When you looked at the actual electron density, what, exactly, did you see that suggested the structure may not be correct? - The structure factors were deposited, so it should be easy for you to try re-refining it yourself with newer software, and see if it makes much difference to the final model. - If you've gone through all those steps, and are still convinced that this structure is problematic, please do let the community know! ++ Phoebe A. Rice Dept. of Biochemistry Molecular Biology The University of Chicago pr...@uchicago.edumailto:pr...@uchicago.edu From: CCP4 bulletin board [CCP4BB@JISCMAIL.AC.UK] on behalf of Misbah ud Din Ahmad [misba.ah...@gmail.com] Sent: Thursday, April 23, 2015 11:03 AM To: CCP4BB@JISCMAIL.AC.UKmailto:CCP4BB@JISCMAIL.AC.UK Subject: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers! Dear crystallographers, The PDB entry http://www.rcsb.org/pdb/explore.do?structureId=3BDN has 16.5% Ramachandran outliers. When I opened this PDB file in coot and checked for Ramachandran outliers, the results are: In preffered region: 58.04% In allowed regions: 19.78% Outliers: 22.17% ! With an R-free of 37.4% at 3.9 A resolution, could you please tell me how reliable this structure of Lambda repressor bound to DNA is? Thanks Misbha Notice: This e-mail message, together with any attachments, contains information of Merck Co., Inc. (2000 Galloping Hill Road, Kenilworth, New Jersey, USA 07033), and/or its affiliates Direct contact information for affiliates is available at http://www.merck.com/contact/contacts.html) that may be confidential, proprietary copyrighted and/or legally privileged. It is intended solely for the use of the individual or entity named on this message. If you are not the intended recipient, and have received this message in error, please notify us immediately by reply e-mail and then delete it from your system.
Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers!!!!!
The abstract of the papers says they used MIR. Mark J van Raaij Dpto de Estructura de Macromoleculas Centro Nacional de Biotecnologia - CSIC c/Darwin 3 E-28049 Madrid, Spain tel. (+34) 91 585 4616 http://www.cnb.csic.es/~mjvanraaij On 23 Apr 2015, at 18:57, Todd Jason Green wrote: My guess is they had the best data they could get, did molecular replacement with the two halves of the repressor and the dna, got a solution and didn't use appropriate restraints in the refinement. Like Phoebe mentioned, we have better tools for this these days. From: CCP4 bulletin board [CCP4BB@JISCMAIL.AC.UK] on behalf of Mark J van Raaij [mjvanra...@cnb.csic.es] Sent: Thursday, April 23, 2015 11:49 AM To: CCP4BB@JISCMAIL.AC.UK Subject: Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers! How reliable is too general a question - it depends on what you want to know. At 3.9Å they could probably place the phosphate atoms quite well and see the general fold of the protein. Finer details will be less reliable, i.e. where the exact side-chains are etc. They could probably have forced more amino acids into favourable Ramachandran angles, but would that have made the structure better? Would these favourable angles have been more right? At 3.9Å you can't know for sure. Would they have been able to draw more biological conclusions? I'd say not. As long as they do not draw more conclusions in the paper than what is supported by the medium-resolution data, the structure provides useful information. Mark J van Raaij Dpto de Estructura de Macromoleculas Centro Nacional de Biotecnologia - CSIC c/Darwin 3 E-28049 Madrid, Spain tel. (+34) 91 585 4616 http://www.cnb.csic.es/~mjvanraaij On 23 Apr 2015, at 18:03, Misbah ud Din Ahmad wrote: Dear crystallographers, The PDB entry http://www.rcsb.org/pdb/explore.do?structureId=3BDN has 16.5% Ramachandran outliers. When I opened this PDB file in coot and checked for Ramachandran outliers, the results are: In preffered region: 58.04% In allowed regions: 19.78% Outliers: 22.17% ! With an R-free of 37.4% at 3.9 A resolution, could you please tell me how reliable this structure of Lambda repressor bound to DNA is? Thanks Misbha
Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers!!!!!
At around 4.0 A resolution one normally cannot talk about accuracy. The density will at most locations not warrant any detailed interpretation. If, at 4.0 A resolution, you move atoms around a bit, you will not see significant changes in R/Rfree. So, you can do whatever you want, more or less. If the refinement puts great emphasis on the secondary structure (i.e. tries to force the Ramachandran plot, then you get a good Ramachandran plot, but if they put more weight of the fit to the density, you get a slightly lower R (and perhaps even Rfree) at the cost of the Ramachandran plot. And all of that is meaningless. At this resolution you have to check if the fold is likely to be correct. I superposed just any high(er) resolution domain with high sequence similarity at the corresponding 3bdn domain, and I take any bet that the fold of 3bdn is right (at least of the domain I looked at, and I extrapolate to the other domain). PDB_REDO (http://www.cmbi.ru.nl/pdb_redo/) cannot improve this one, and WHAT_CHECK (http://swift.cmbi.ru.nl/gv/pdbreport/) throws its hands up in the air. At 4.0 A you should be happy with an indication of the fold (and with the fact that the authors probably went through great pains for you getting these coordinates close to where they should be). Gert
Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers!!!!!
Hi, I entirely agree with Phoebe. This is 2008 structure (publish date, so refinement was probably done in 2007) - that is before all the new methods and tools for low-res refinement became available. Now, let's have a closer look... Given rather large amount of various geometry outliers (in model taken straight from PDB): Ramachandran plot, number of: outliers : 76(16.52 %) allowed : 120 (26.09 %) favored : 264 (57.39 %) Rotamer outliers: 60 (15.27 %) Cbeta deviations 0.25A : 6 All-atom clashscore : 53.54 it is unlikely for conventional gradient-driven refinement tools to correct them (since correcting some of them requires jumping over energy barriers). Now, if I idealize the model geometry using 2mFo-DFc map as restraint: phenix.real_space_refine 3bd.pdb map.ccp4 run=minimization_global+morphing morphing=every_macro_cycle I get the model that still fits the map as well as staring model (starting map CC vs final map CC: 0.737/0.782) but has outliers virtually gone: Ramachandran plot: outliers : 1.52 % allowed : 19.35 % favored : 79.13 % Rotamer outliers : 0.00 % Cbeta deviations : 0 All-atom clashscore : 17.09 Now if I run this model through reciprocal space refinement so that it fits the Fobs data under restraints to preserve good input geometry (phenix.refine with rotamer and secondary-structure restraints, weights optimization), I get the model that still has reasonable geometry: Ramachandran plot: outliers : 1.74 % allowed : 20.22 % favored : 78.04 % Rotamer outliers : 0.00 % Cbeta deviations : 0 All-atom clashscore : 18.16 with R-factors R-work = 0.3012, R-free = 0.3518 that are slightly better than published (0.292/0.374, taken from PDB). Bottom line in my opinion is: we can't really blame rather old(ish) models for not using modern improved methods, but there is no excuse for not using these methods today! All the best, Pavel On Thu, Apr 23, 2015 at 9:03 AM, Misbah ud Din Ahmad misba.ah...@gmail.com wrote: Dear crystallographers, The PDB entry http://www.rcsb.org/pdb/explore.do?structureId=3BDN has 16.5% Ramachandran outliers. When I opened this PDB file in coot and checked for Ramachandran outliers, the results are: In preffered region: 58.04% In allowed regions: 19.78% *Outliers: 22.17% !* With an R-free of 37.4% at 3.9 A resolution, could you please tell me how reliable this structure of Lambda repressor bound to DNA is?
Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers!!!!!
I feel rather uncomfortable and cornered with the aggressive replies and phrases like throwing stones, public humiliation etc., which suggest that I have some personal enmity with the author. The structure is in the public domain and questions can be asked about it. I as a beginner in crystallography learn from each discussion on this board and such harsh and insinuating responses may, in future, deter people like me from posting any questions at all and learning refinement strategies like the one which Pavel outlined to improve such type of structures. Best Misbha On Thu, Apr 23, 2015 at 9:17 PM, Gert Vriend gerrit.vri...@radboudumc.nl wrote: At around 4.0 A resolution one normally cannot talk about accuracy. The density will at most locations not warrant any detailed interpretation. If, at 4.0 A resolution, you move atoms around a bit, you will not see significant changes in R/Rfree. So, you can do whatever you want, more or less. If the refinement puts great emphasis on the secondary structure (i.e. tries to force the Ramachandran plot, then you get a good Ramachandran plot, but if they put more weight of the fit to the density, you get a slightly lower R (and perhaps even Rfree) at the cost of the Ramachandran plot. And all of that is meaningless. At this resolution you have to check if the fold is likely to be correct. I superposed just any high(er) resolution domain with high sequence similarity at the corresponding 3bdn domain, and I take any bet that the fold of 3bdn is right (at least of the domain I looked at, and I extrapolate to the other domain). PDB_REDO (http://www.cmbi.ru.nl/pdb_redo/) cannot improve this one, and WHAT_CHECK (http://swift.cmbi.ru.nl/gv/pdbreport/) throws its hands up in the air. At 4.0 A you should be happy with an indication of the fold (and with the fact that the authors probably went through great pains for you getting these coordinates close to where they should be). Gert
Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers!!!!!
Well, with full respect to your sensitivities as far as your own person is concerned: if you play rough (and 5 exclamation marks qualify by commonly accepted email etiquette as incipient flame, at least), you need to be willing to take a few as well... Best, BR - Trigger warning --- This message may or may not contain references to issues of privilege and oppression in- cluding but not limited to enantiophobia, point classism, heteroatomism, transbondism, cisbonding, sizeism, curvature, references to color, alcohol, blood, small insects, cancer cells, and any combination thereof that may cause symptoms ranging from discomfort and anxiety to violent physical response in sensitive individuals and must therefore be labeled as potentially hazardous to your comfort and well-being. - Trigger warning --- From: CCP4 bulletin board [mailto:CCP4BB@JISCMAIL.AC.UK] On Behalf Of Misbah ud Din Ahmad Sent: Thursday, April 23, 2015 2:05 PM To: CCP4BB@JISCMAIL.AC.UK Subject: Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers! I feel rather uncomfortable and cornered with the aggressive replies and phrases like throwing stones, public humiliation etc., which suggest that I have some personal enmity with the author. The structure is in the public domain and questions can be asked about it. I as a beginner in crystallography learn from each discussion on this board and such harsh and insinuating responses may, in future, deter people like me from posting any questions at all and learning refinement strategies like the one which Pavel outlined to improve such type of structures. Best Misbha On Thu, Apr 23, 2015 at 9:17 PM, Gert Vriend gerrit.vri...@radboudumc.nl wrote: At around 4.0 A resolution one normally cannot talk about accuracy. The density will at most locations not warrant any detailed interpretation. If, at 4.0 A resolution, you move atoms around a bit, you will not see significant changes in R/Rfree. So, you can do whatever you want, more or less. If the refinement puts great emphasis on the secondary structure (i.e. tries to force the Ramachandran plot, then you get a good Ramachandran plot, but if they put more weight of the fit to the density, you get a slightly lower R (and perhaps even Rfree) at the cost of the Ramachandran plot. And all of that is meaningless. At this resolution you have to check if the fold is likely to be correct. I superposed just any high(er) resolution domain with high sequence similarity at the corresponding 3bdn domain, and I take any bet that the fold of 3bdn is right (at least of the domain I looked at, and I extrapolate to the other domain). PDB_REDO (http://www.cmbi.ru.nl/pdb_redo/) cannot improve this one, and WHAT_CHECK (http://swift.cmbi.ru.nl/gv/pdbreport/) throws its hands up in the air. At 4.0 A you should be happy with an indication of the fold (and with the fact that the authors probably went through great pains for you getting these coordinates close to where they should be). Gert
Re: [ccp4bb] 3BDN, 16.5% Ramachandran Outliers!!!!!
How reliable is too general a question - it depends on what you want to know. At 3.9Å they could probably place the phosphate atoms quite well and see the general fold of the protein. Finer details will be less reliable, i.e. where the exact side-chains are etc. They could probably have forced more amino acids into favourable Ramachandran angles, but would that have made the structure better? Would these favourable angles have been more right? At 3.9Å you can't know for sure. Would they have been able to draw more biological conclusions? I'd say not. As long as they do not draw more conclusions in the paper than what is supported by the medium-resolution data, the structure provides useful information. Mark J van Raaij Dpto de Estructura de Macromoleculas Centro Nacional de Biotecnologia - CSIC c/Darwin 3 E-28049 Madrid, Spain tel. (+34) 91 585 4616 http://www.cnb.csic.es/~mjvanraaij On 23 Apr 2015, at 18:03, Misbah ud Din Ahmad wrote: Dear crystallographers, The PDB entry http://www.rcsb.org/pdb/explore.do?structureId=3BDN has 16.5% Ramachandran outliers. When I opened this PDB file in coot and checked for Ramachandran outliers, the results are: In preffered region: 58.04% In allowed regions: 19.78% Outliers: 22.17% ! With an R-free of 37.4% at 3.9 A resolution, could you please tell me how reliable this structure of Lambda repressor bound to DNA is? Thanks Misbha
