[ccp4bb] geometry problems with sugars
Hi All, My question is concerning geometry of NAGs in a glycoprotein structure. I recently solved the structure of a glycoprotein to 3 Å and modeled NAGs linked to Asn at 3 different places. NAGs and Asn-NAG links are refined in Phenix.refine as per the Phenix dictionary. However, when submitting the structure to the PDB, internal validation of PDB found that NAGs at 2 places have geometry problems (atoms surrounding the C1 of NAG are in the same plane). What I learned from Phenix bulletin board is that the refinement program is probably fixing the NAG into a local minimal structure to fit to the density the best (I have OK density for sugars) and that's causing the problem. So, I tried to fix the geometry of NAG while refining, so that the refinement does not change the geometry of the sugar. But, still the internal validation of PDB found the same problem with the sugars. Then, I tried replacing the NAGs with ideal monomers from Coot. Still, the problem persisted. PDB annotator's suggestion is to get the NAG coordinates from HIC-Up and try refinement in another program. I am wondering if any one else noticed (or had ) a similar problem with NAG geometry using either Phenix.refine or Coot. What baffles me is that the ideal NAG from coot dictionary did not pass the internal validation of the PDB. I would appreciate if any one has any suggestion (other than trying a different refinement program) to get around this problem. Is there a way to compare the NAGs in your structure to the ideal and get to know what to fix ? Thanks in advance Tirumal
Re: [ccp4bb] geometry problems with sugars
tirumal wrote: Hi All, My question is concerning geometry of NAGs in a glycoprotein structure. Fire away... I recently solved the structure of a glycoprotein to 3 Å and modeled NAGs linked to Asn at 3 different places. NAGs and Asn-NAG links are refined in Phenix.refine as per the Phenix dictionary. Errr.. that'll be the Refmac dictionary. However, when submitting the structure to the PDB, internal validation of PDB found that NAGs at 2 places have geometry problems (atoms surrounding the C1 of NAG are in the same plane). The surrounding atoms being the ND2, O5 and C2, I presume. The C1 should not (of course) be in the plane of those 3 atoms. What I learned from Phenix bulletin board is that the refinement program is probably fixing the NAG into a local minimal structure to fit to the density the best (I have OK density for sugars) and that's causing the problem. It seems unlikely to me that you density would be sufficiently strong to put C1, ND2, O5 and C2 into a plane. So, I tried to fix the geometry of NAG while refining, so that the refinement does not change the geometry of the sugar. But, still the internal validation of PDB found the same problem with the sugars. I don't follow this. How can you fix the atoms of the NAG and then refine it (and expect things to move)? Then, I tried replacing the NAGs with ideal monomers from Coot. "Get Monomer" in Coot use LIBCHECK to generate coordinates and restraints. I can use "Get Monomer" to import a NAG and create and refine a N-link quite happily (after deleting the O1 and HO1 of course - Coot won't do that for you yet). Still, the problem persisted. PDB annotator's suggestion is to get the NAG coordinates from HIC-Up and try refinement in another program. A very worthy path to pursue... I am wondering if any one else noticed (or had ) a similar problem with NAG geometry using either Phenix.refine or Coot. I don't see a problem. What baffles me is that the ideal NAG from coot dictionary did not pass the internal validation of the PDB. I'd find that surprising too (and to be accurate, it's from LIBCHECK and just imported into Coot). I would appreciate if any one has any suggestion (other than trying a different refinement program) to get around this problem. Perish the thought of suggesting a CCP4 refinement program on this list... :) Is there a way to compare the NAGs in your structure to the ideal and get to know what to fix ? I must admit that I'm curious to know what the PDB has against CCP4's NAG. HTH, Paul.
Re: [ccp4bb] geometry problems with sugars
I've seen flattening of the C1 atom recently in Coot - in that case the reason was that, as the density wasn't really clear around the sugar, I'd fitted it 180 degrees rotated around the ASN-ND2---C1-NAG bond. In such cases Coot happily flattens the sugar ring such that C1 C2 C3 C5 O5 are in a plane, just locked half way between chair and boat. Bug? Or yet another bite of the torsions/chiralities monster? -- AstraZeneca UK Limited is a company incorporated in England and Wales with registered number: 03674842 and a registered office at 15 Stanhope Gate, London W1K 1LN. Confidentiality Notice: This message is private and may contain confidential, proprietary and legally privileged 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 unauthorised use or disclosure of the contents of this message is not permitted and may be unlawful. Disclaimer: Email messages may be subject to delays, interception, non-delivery and unauthorised alterations. Therefore, information expressed in this message is not given or endorsed by AstraZeneca UK Limited unless otherwise notified by an authorised representative independent of this message. No contractual relationship is created by this message by any person unless specifically indicated by agreement in writing other than email. Monitoring: AstraZeneca UK Limited may monitor email traffic data and content for the purposes of the prevention and detection of crime, ensuring the security of our computer systems and checking Compliance with our Code of Conduct and Policies. -Original Message- From: CCP4 bulletin board [mailto:ccp...@jiscmail.ac.uk] On Behalf Of Paul Emsley Sent: 20 April 2010 23:57 To: CCP4BB@JISCMAIL.AC.UK Subject: Re: [ccp4bb] geometry problems with sugars tirumal wrote: > Hi All, > > My question is concerning geometry of NAGs in a glycoprotein structure. > Fire away... > > I recently solved the structure of a glycoprotein to 3 Å and modeled > NAGs linked to Asn at 3 different places. NAGs and Asn-NAG links are > refined in Phenix.refine as per the Phenix dictionary. > Errr.. that'll be the Refmac dictionary. > > However, when submitting the structure to the PDB, internal validation > of PDB found that NAGs at 2 places have geometry problems (atoms > surrounding the C1 of NAG are in the same plane). > The surrounding atoms being the ND2, O5 and C2, I presume. The C1 should not (of course) be in the plane of those 3 atoms. > What I learned from Phenix bulletin board is that the refinement > program is probably fixing the NAG into a local minimal structure to > fit to the density the best (I have OK density for sugars) and that's > causing the problem. > It seems unlikely to me that you density would be sufficiently strong to put C1, ND2, O5 and C2 into a plane. > > So, I tried to fix the geometry of NAG while refining, so that the > refinement does not change the geometry of the sugar. But, still the > internal validation of PDB found the same problem with the sugars. > I don't follow this. How can you fix the atoms of the NAG and then refine it (and expect things to move)? > > Then, I tried replacing the NAGs with ideal monomers from Coot. > "Get Monomer" in Coot use LIBCHECK to generate coordinates and restraints. I can use "Get Monomer" to import a NAG and create and refine a N-link quite happily (after deleting the O1 and HO1 of course - Coot won't do that for you yet). > Still, the problem persisted. PDB annotator's suggestion is to get the > NAG coordinates from HIC-Up and try refinement in another program. > A very worthy path to pursue... > > I am wondering if any one else noticed (or had ) a similar problem > with NAG geometry using either Phenix.refine or Coot. > I don't see a problem. > What baffles me is that the ideal NAG from coot dictionary did not > pass the internal validation of the PDB. > I'd find that surprising too (and to be accurate, it's from LIBCHECK and just imported into Coot). > > I would appreciate if any one has any suggestion (other than trying a > different refinement program) to get around this problem. > Perish the thought of suggesting a CCP4 refinement program on this list... :) > Is there a way to compare the NAGs in your structure to the ideal and > get to know what to fix ? > > I must admit that I'm curious to know what the PDB has against CCP4's NAG. HTH, Paul.
Re: [ccp4bb] geometry problems with sugars
As I see there is no chirality definition for NAG-ASN link (perhaps there should be but then people will be unhappy even more). Only reason i can see for this flattening is conflict between geometry and electron density. Your example shows that even if electron density is weak it may play a role and correct orientation of sugar may matter. And there is no torsion angle restraint for this particular link (at least no dictionary definition for this. Programs may generate them on fly but I would not do it) As I see it is not a bug or infamous chirality problem (poor chirality: it is getting attacked from all four corners) but shortcomings of fitting into electron density (user or program, it is a different matter). Full link description (apart from removal of O1 HO1 of and HD22 of ASN) is here: data_link_NAG-ASN # loop_ _chem_link_bond.link_id _chem_link_bond.atom_1_comp_id _chem_link_bond.atom_id_1 _chem_link_bond.atom_2_comp_id _chem_link_bond.atom_id_2 _chem_link_bond.type _chem_link_bond.value_dist _chem_link_bond.value_dist_esd NAG-ASN 1 C1 2 ND2 single 1.439 .020 loop_ _chem_link_angle.link_id _chem_link_angle.atom_1_comp_id _chem_link_angle.atom_id_1 _chem_link_angle.atom_2_comp_id _chem_link_angle.atom_id_2 _chem_link_angle.atom_3_comp_id _chem_link_angle.atom_id_3 _chem_link_angle.value_angle _chem_link_angle.value_angle_esd NAG-ASN 1 C1 2 ND2 2 CG 121.0003.000 NAG-ASN 1 C1 2 ND2 2 HD2 119.0003.000 NAG-ASN 1 O5 1 C1 2 ND2 112.3003.000 loop_ _chem_link_plane.link_id _chem_link_plane.plane_id _chem_link_plane.atom_comp_id _chem_link_plane.atom_id _chem_link_plane.dist_esd NAG-ASNplane11 C1 .020 NAG-ASNplane12 ND2.020 NAG-ASNplane12 CG .020 NAG-ASNplane12 OD1.020 NAG-ASNplane12 CB .020 NAG-ASNplane12 HD21 .020 Regards Garib On 21 Apr 2010, at 10:37, Debreczeni, Judit wrote: I've seen flattening of the C1 atom recently in Coot - in that case the reason was that, as the density wasn't really clear around the sugar, I'd fitted it 180 degrees rotated around the ASN-ND2---C1-NAG bond. In such cases Coot happily flattens the sugar ring such that C1 C2 C3 C5 O5 are in a plane, just locked half way between chair and boat. Bug? Or yet another bite of the torsions/chiralities monster? -- AstraZeneca UK Limited is a company incorporated in England and Wales with registered number: 03674842 and a registered office at 15 Stanhope Gate, London W1K 1LN. Confidentiality Notice: This message is private and may contain confidential, proprietary and legally privileged 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 unauthorised use or disclosure of the contents of this message is not permitted and may be unlawful. Disclaimer: Email messages may be subject to delays, interception, non-delivery and unauthorised alterations. Therefore, information expressed in this message is not given or endorsed by AstraZeneca UK Limited unless otherwise notified by an authorised representative independent of this message. No contractual relationship is created by this message by any person unless specifically indicated by agreement in writing other than email. Monitoring: AstraZeneca UK Limited may monitor email traffic data and content for the purposes of the prevention and detection of crime, ensuring the security of our computer systems and checking Compliance with our Code of Conduct and Policies. -Original Message- From: CCP4 bulletin board [mailto:ccp...@jiscmail.ac.uk] On Behalf Of Paul Emsley Sent: 20 April 2010 23:57 To: CCP4BB@JISCMAIL.AC.UK Subject: Re: [ccp4bb] geometry problems with sugars tirumal wrote: Hi All, My question is concerning geometry of NAGs in a glycoprotein structure. Fire away... I recently solved the structure of a glycoprotein to 3 Å and modeled NAGs linked to Asn at 3 different places. NAGs and Asn-NAG links are refined in Phenix.refine as per the Phenix dictionary. Errr.. that'll be the Refmac dictionary. However, when submitting the structure to the PDB, internal validation of PDB found that NAGs at 2 places have geometry problems (atoms surrounding the C1 of NAG are in the same plane). The surrounding atoms being the ND2, O5 and C2, I presume. The C1 should not (of course) be in the plane of those 3 atoms. What I learned from Phenix bulletin board is that the refinement program is probably fixing the NAG into a local minimal structure to fit to the density the best (I have OK density for sugars) and that's causing the problem. It seems unlikely to me that you dens
Re: [ccp4bb] geometry problems with sugars
Garib Murshudov wrote: As I see there is no chirality definition for NAG-ASN link (perhaps there should be but then people will be unhappy even more). Only reason i can see for this flattening is conflict between geometry and electron density. Your example shows that even if electron density is weak it may play a role and correct orientation of sugar may matter. I agree, and with JED too. More tests suggest that if I put the NAG into the density the wrong way round, Coot will happily flatten the C1. So, my guess would be that if you rotated your NAG 180 degrees round a vector ~ NG--(midpoint of C3,C4) and re-refined, then things would improve. At the moment, there is no substitute for knowledge when building carbohydrates - it would be a substantial improvement I think if someone added intelligent carbohydrate validation tools into Coot. Paul.
Re: [ccp4bb] geometry problems with sugars
JED's example is very illustrative and it shows that chirality may need to be added to this link definition. then sugar validation may be easier (at least ASN-NAG with only one sugar). If chirality is wrong then rotate around ND2-C1bond as a rigid group. Just like you do with rotamers. Here you have only two orientations. Garib On 21 Apr 2010, at 14:20, Paul Emsley wrote: Garib Murshudov wrote: As I see there is no chirality definition for NAG-ASN link (perhaps there should be but then people will be unhappy even more). Only reason i can see for this flattening is conflict between geometry and electron density. Your example shows that even if electron density is weak it may play a role and correct orientation of sugar may matter. I agree, and with JED too. More tests suggest that if I put the NAG into the density the wrong way round, Coot will happily flatten the C1. So, my guess would be that if you rotated your NAG 180 degrees round a vector ~ NG--(midpoint of C3,C4) and re-refined, then things would improve. At the moment, there is no substitute for knowledge when building carbohydrates - it would be a substantial improvement I think if someone added intelligent carbohydrate validation tools into Coot. Paul.
Re: [ccp4bb] geometry problems with sugars
Thanks to all who responded. 180 degrees flip of the problematic NAGs, did help. > At the moment, there is no substitute for knowledge when building carbohydrates - it >would be a substantial improvement I think if someone added intelligent carbohydrate >validation tools into Coot. If you have a poor density (which I guess, generally is the case for large glycoprotein structures) you have to depend on trial and error strategy to get the right NAG conformation. I don't know how other refinement programs handle this, but after Phenix.refinement run, one has to definitely check the geometry of the NAGs carefully. Hope to see a validation tool for NAGs in Coot soon. Tirumal --- On Wed, 21/4/10, Garib Murshudov wrote: From: Garib Murshudov Subject: Re: [ccp4bb] geometry problems with sugars To: CCP4BB@JISCMAIL.AC.UK Date: Wednesday, 21 April, 2010, 9:58 JED's example is very illustrative and it shows that chirality may need to be added to this link definition. then sugar validation may be easier (at least ASN-NAG with only one sugar). If chirality is wrong then rotate around ND2-C1bond as a rigid group. Just like you do with rotamers. Here you have only two orientations. Garib On 21 Apr 2010, at 14:20, Paul Emsley wrote: > Garib Murshudov wrote: >> As I see there is no chirality definition for NAG-ASN link (perhaps there >> should be but then people will be unhappy even more). >> Only reason i can see for this flattening is conflict between geometry and >> electron density. Your example shows that even if electron density is weak >> it may play a role and correct orientation of sugar may matter. >> > > I agree, and with JED too. More tests suggest that if I put the NAG into the > density the wrong way round, Coot will happily flatten the C1. So, my guess > would be that if you rotated your NAG 180 degrees round a vector ~ > NG--(midpoint of C3,C4) and re-refined, then things would improve. > > At the moment, there is no substitute for knowledge when building > carbohydrates - it would be a substantial improvement I think if someone > added intelligent carbohydrate validation tools into Coot. > > Paul.
Re: [ccp4bb] geometry problems with sugars
Dear sugar loving people, I have been working with a lot of glycoproteins (up to 30 % carbohydrates) at resolutions as "bad" as 2.8 Å. Nevertheless, I was able to built sometimes about 10 sugar moieties/carbohydrate chain. Although, sugar molecules usually have a somewhat bulky density, and don´t have such distinct structural features like in proteins (e.g. main chain, side chain, C=O bump) one can use geometric restraints to place the sugars correctly with a high probability. I other words I usually first superimpose the e.g. O1 of the sugar molecule with the N of the Asn or O of the Ser/Thr or with the e.g. O4 of the previous sugar (looking from Asn/Ser/Thr). Then I just rotate about the glycosidic bond or only about the O1 (e.g. C1-O4 for a beta1-4) and try to fit the density as good as possible. Often real space refinement in COOT helps after this stage to optimize. Then I check the interactions of the newly placed sugar with its close neighbourhood. Well defined sugars, and only those you can see in a crystal structure, make practically with each of their OH groups or any other polar atom/group (e.g. the N in NAG) at least one H-bond, either with the protein, with other neighbouring sugars, or with waters or even with other solvent molecules like SO4. Then you should also check reasonable van der Waals distances of non-polar atoms. If you find a conformation where you have the optimum number of H-bonds and no outliers (e. g. < 3.2 Å) of close non-bonding contacts, you most likely have placed your sugar molecule correctly. In e.g. REFMAC with the review mode you can then check if you have a alpha or beta glycosidic bond and if this is what you expect. BR, Klaus Am 22.04.2010 um 19:13 schrieb tirumal: Thanks to all who responded. 180 degrees flip of the problematic NAGs, did help. > At the moment, there is no substitute for knowledge when building carbohydrates - it >would be a substantial improvement I think if someone added intelligent carbohydrate >validation tools into Coot. If you have a poor density (which I guess, generally is the case for large glycoprotein structures) you have to depend on trial and error strategy to get the right NAG conformation. I don't know how other refinement programs handle this, but after Phenix.refinement run, one has to definitely check the geometry of the NAGs carefully. Hope to see a validation tool for NAGs in Coot soon. Tirumal --- On Wed, 21/4/10, Garib Murshudov wrote: From: Garib Murshudov Subject: Re: [ccp4bb] geometry problems with sugars To: CCP4BB@JISCMAIL.AC.UK Date: Wednesday, 21 April, 2010, 9:58 JED's example is very illustrative and it shows that chirality may need to be added to this link definition. then sugar validation may be easier (at least ASN-NAG with only one sugar). If chirality is wrong then rotate around ND2-C1bond as a rigid group. Just like you do with rotamers. Here you have only two orientations. Garib On 21 Apr 2010, at 14:20, Paul Emsley wrote: > Garib Murshudov wrote: >> As I see there is no chirality definition for NAG-ASN link (perhaps there should be but then people will be unhappy even more). >> Only reason i can see for this flattening is conflict between geometry and electron density. Your example shows that even if electron density is weak it may play a role and correct orientation of sugar may matter. >> > > I agree, and with JED too. More tests suggest that if I put the NAG into the density the wrong way round, Coot will happily flatten the C1. So, my guess would be that if you rotated your NAG 180 degrees round a vector ~ NG--(midpoint of C3,C4) and re-refined, then things would improve. > > At the moment, there is no substitute for knowledge when building carbohydrates - it would be a substantial improvement I think if someone added intelligent carbohydrate validation tools into Coot. > > Paul. Dr. Klaus Piontek Albert-Ludwigs-University Freiburg Institute of Organic Chemistry and Biochemistry, Room 401 H Albertstrasse 21 D-79104 Freiburg Germany Phone: ++49-761-203-6036 Fax: ++49-761-203-8714 Email: klaus.pion...@ocbc.uni-freiburg.de Web: http://www.chemie.uni-freiburg.de/orgbio/w3platt/
Re: [ccp4bb] geometry problems with sugars
I am trying to resist answering ccp4bb ... but > Thanks to all who responded. 180 degrees flip of the problematic NAGs, > did help. If you have a poor density (which I guess, generally is the > case for large glycoprotein structures) you have to depend on trial > and error strategy to get the right NAG conformation. I don't know > how other refinement programs handle this, but after Phenix.refinement > run, one has to definitely check the geometry of the NAGs carefully. There is no such thing as 'trial-and-error' Look up the polysaccharide you are expecting to find for the species you are working with see for example Varki A et al. Essentials of Glycobiology. Cold Spring Harbor Laboratory Press; 2nd edition http://www.ncbi.nlm.nih.gov/bookshelf/br.fcgi?book=glyco2 Note fundamentally that an Asn-Nag is always beta- so why not check in the first place that you have NAG attached in a beta conformation if you have alpha - then you are wrong and the density will flatten the C1 - The ccp4bb notes about rotating 180deg is misleading as it is not a conformational change it is a configurational change. It is made by an enzyme not by a crystallographer - all chemistry in biology is run by enzymes each species makes particular sugars in a particular order look it up, you cant get a oligosaccharide that cant be made by a particular species You are not going to find an N-linked Glycan unknown to glycobiology so first of all look up what is expected e.g. http://www.glycoforum.gr.jp/ http://www.genome.jp/ligand/kcam/ http://www.functionalglycomics.org/static/index.shtml http://www.glyco.ac.ru/bcsdb3/ http://www.casper.organ.su.se/ECODAB/ http://www.functionalglycomics.org/static/gt/gtdb.shtml http://akashia.sci.hokudai.ac.jp/ http://hexose.chem.ku.edu/sugar.php http://www.eurocarbdb.org/ then try http://www.glycosciences.de/modeling/sweet2/doc/index.php this will give you a 3D model in PDB format of your maximum glycan - dont do what has happened in the past in the PDB just try adding sugars at random connections and configurations you can also look at http://www.ebi.ac.uk/eurocarb/gwb/builder.action There are tools on http://www.glycosciences.de/ e.g. http://www.glycosciences.de/modeling/glycomapsdb/ to tell you the expected conformational maps of the glycosidic linkage - you believe in Ramachandran for peptide bonds so why not for glycosidic bonds) The PDB has a legacy of wrong connections, wrong conformations, poor geometry of oligosaccharides especially glycans where the coordinates respresent unknown molecules The density may be poor but if you know what say a plant may produce then fit what you can of that particular saccharide in observable density the linkages, 1-2, 1-3, 1-4, 1-6 are all specified and people should stop making them up at random for example plants have basically these N-glycans http://www.ncbi.nlm.nih.gov/bookshelf/br.fcgi?book=glyco2&part=ch22&rendertype=figure&id=ch22.f1 sweet (above) will make the 3d coordinates you can try fitting as much as is observed knowing the configuration and trying the conformations from glycomapsdb
Re: [ccp4bb] geometry problems with sugars
On Thu, 22 Apr 2010 10:13:40 -0700, tirumal wrote: > If you have a poor density (which I guess, generally is the case for large > glycoprotein structures) you have to depend on trial and error strategy to > get the right NAG conformation. I don't know how other refinement programs > handle this, but after Phenix.refinement run, one has to definitely check the > geometry of the NAGs carefully. I would like to add two further resources to the list of URLs already posted by Kim in this thread: You can use the PDB CArbohydrate REsidue check (pdb-care) tool to check carbohydrate residue geometry: http://www.glycosciences.de/tools/pdb-care/ This program detects carbohydrates, tries and assigns correct PDB residue names and checks, if they match the residue names used in the uploaded file. We will also add some further validation of N-glycan structures to notify users of "unusual" residues such as a-D-GlcpNAc (NDG) instead of b-D-GlcpNAc (NAG) in the N-glycan core structure. Kim already mentioned GlycoMapsDB to find the preferred conformations. The CARP (CArbohydrate Ramachandran Plot, http://www.glycosciences.de/tools/carp/) software helps you to check the conformation of glycosidic linkages by creating Ramachandran plot-like images. You can compare the glycosidic torsions in your structure either to computed maps from the GlycoMapsDB or to the glycosidic linkage conformations found in the PDB. > Hope to see a validation tool for NAGs in Coot soon. We are planning to implement some of the pdb-care checks in Coot, but of course this will take some time. Thomas
