Re: [COOT] CHO H-bonds.
About the unit cell shrinkage: I have heard a lecture (I believe it was Andrea Thorn form G. Sheldrick group on one of CCP4 meetings), where she mentioned a mistake in X-ray wavelenght input (swaped 3rd and 4th digit), which no program from integration to structure solution did not recognised. It appeared in structure refinement in wrong bond distances. Worth of checking. Jan On 07/01/2015 10:58 AM, Ian Tickle wrote: Paul, as an aside to this, would it be possible to have it so that if the relevant boxes in the Environment Distances menu are checked then H-bonds and.or bumps are automatically re-calculated on re-centering? Currently it's necessary to check off on one of the boxes every time the view is re-centered. If I forget to do that it's very easy to miss bumps (and I don't have a great deal of faith in MP's concept of what constitutes a bump)! Cheers -- Ian On 30 June 2015 at 14:33, Ian Tickle ianj...@gmail.com mailto:ianj...@gmail.com wrote: Hello All I guess this is really a question about MolProbity (and possibly about autoBuster) but I assume that most Coot users will be using the MolProbity validation tools. I am in the process of depositing 4 structures of the same protein (different ligands) and I noticed that MP seems to be reporting an unusually large number of bumps in both the small overlap and bad overlap classes. In each case the resolution is 2 Ang., the structures have been refined by autoBuster and the density seems to be unequivocal, see e.g.: https://drive.google.com/file/d/0B4H4H-DyO60-SEQwS1k5S3RIVG8/view?usp=sharing A lot of the bumps are main-chain CHalpha to main-chain carbonyl O H-bonds, but there are also some CH...O side-chain H-bonds, again with clear density. The C...O distances are in the range 3.0 to 3.2 Ang., so too short for a vdW contact. The H...O distances are ~ 2.2 Ang. which is definitely shorter than the sum of the vdW radii ( H: 1.2 + O: 1.5 = 2.7). I found this survey of CH...O H-bonds but it's restricted to CH...O bonds at the end of helices and I see them mostly in sheets. http://www.mrc-lmb.cam.ac.uk/genomes/madanm/pdfs/chapter1.pdf This reports 11 examples (i.e. H-bonds, not structures) in the 3.0-3.2 range for the whole of the PDB (admittedly as it was in 2001 when the article appears to have been written). I have about the same number in one structure! One possibility I considered was that the unit cells had all somehow 'shrunk'. This can be tested with WhatCheck: however it only reports a very small shrinkage which translates to an error of ~ 0.02 Ang. in a 3 Ang. distance, which is nowhere near enough to explain a discrepancy of 0.5 Ang. So I guess my question is has anyone else noticed this in their MP dot-plots; also does anyone know what criteria does MP uses for testing bumps and specifically what value is it using for CH...O H-bonds? And of course I'd like to know whether this will affect my percentile ranking in the clashscore from the PDB validation! Cheers -- Ian -- Jan Stransky, PhD student Institute of Biotechnology, CAS Laboratory of structure and function of biomolecules Nad Safinou II 366 Vestec Czech Republic Tel.: +420226201570
Re: [COOT] CHO H-bonds.
Dear Tim, I know that it cause rescaling and you don't have to input the number manually most of the time recently. I just wanted to point out that it can obviously happend. The question reminded me Andrea's lecture. Jan On 07/02/2015 10:34 AM, Tim Gruene wrote: -BEGIN PGP SIGNED MESSAGE- Hash: SHA1 Dear Jan, when you input the wrong wavelength during data processing, you rescale the Ewald sphere construction, i.e. it will result in a different unit cell and crystal to detector distance. I am not sure what you mean by 'which no program from integration to structure solution did not recognised'. How should the downstream software detect the human mistake? The wavelength you enter should of course be double checked. Nowadays this seldomly occurs because the experimental settings are usually transferred automatically and reliably between the beamline software and the processing tools. Cheers, Tim On 07/02/2015 09:27 AM, Jan Stransky wrote: About the unit cell shrinkage: I have heard a lecture (I believe it was Andrea Thorn form G. Sheldrick group on one of CCP4 meetings), where she mentioned a mistake in X-ray wavelenght input (swaped 3rd and 4th digit), which no program from integration to structure solution did not recognised. It appeared in structure refinement in wrong bond distances. Worth of checking. Jan On 07/01/2015 10:58 AM, Ian Tickle wrote: Paul, as an aside to this, would it be possible to have it so that if the relevant boxes in the Environment Distances menu are checked then H-bonds and.or bumps are automatically re-calculated on re-centering? Currently it's necessary to check off on one of the boxes every time the view is re-centered. If I forget to do that it's very easy to miss bumps (and I don't have a great deal of faith in MP's concept of what constitutes a bump)! Cheers -- Ian On 30 June 2015 at 14:33, Ian Tickle ianj...@gmail.com mailto:ianj...@gmail.com wrote: Hello All I guess this is really a question about MolProbity (and possibly about autoBuster) but I assume that most Coot users will be using the MolProbity validation tools. I am in the process of depositing 4 structures of the same protein (different ligands) and I noticed that MP seems to be reporting an unusually large number of bumps in both the small overlap and bad overlap classes. In each case the resolution is 2 Ang., the structures have been refined by autoBuster and the density seems to be unequivocal, see e.g.: https://drive.google.com/file/d/0B4H4H-DyO60-SEQwS1k5S3RIVG8/view?usp=sharing A lot of the bumps are main-chain CHalpha to main-chain carbonyl O H-bonds, but there are also some CH...O side-chain H-bonds, again with clear density. The C...O distances are in the range 3.0 to 3.2 Ang., so too short for a vdW contact. The H...O distances are ~ 2.2 Ang. which is definitely shorter than the sum of the vdW radii ( H: 1.2 + O: 1.5 = 2.7). I found this survey of CH...O H-bonds but it's restricted to CH...O bonds at the end of helices and I see them mostly in sheets. http://www.mrc-lmb.cam.ac.uk/genomes/madanm/pdfs/chapter1.pdf This reports 11 examples (i.e. H-bonds, not structures) in the 3.0-3.2 range for the whole of the PDB (admittedly as it was in 2001 when the article appears to have been written). I have about the same number in one structure! One possibility I considered was that the unit cells had all somehow 'shrunk'. This can be tested with WhatCheck: however it only reports a very small shrinkage which translates to an error of ~ 0.02 Ang. in a 3 Ang. distance, which is nowhere near enough to explain a discrepancy of 0.5 Ang. So I guess my question is has anyone else noticed this in their MP dot-plots; also does anyone know what criteria does MP uses for testing bumps and specifically what value is it using for CH...O H-bonds? And of course I'd like to know whether this will affect my percentile ranking in the clashscore from the PDB validation! Cheers -- Ian - -- - -- Dr Tim Gruene Institut fuer anorganische Chemie Tammannstr. 4 D-37077 Goettingen phone: +49 (0)551 39 22149 GPG Key ID = A46BEE1A -BEGIN PGP SIGNATURE- Version: GnuPG v1 iD8DBQFVlPeFUxlJ7aRr7hoRAuQXAJ4gYxRxL/BELbrDGpAxUXbUErBPgwCghcsv nEVcp0relCBO/HVt9poCnYE= =V1FJ -END PGP SIGNATURE- -- Jan Stransky, PhD student Institute of Biotechnology, CAS Laboratory of structure and function of biomolecules Nad Safinou II 366 Vestec Czech Republic Tel.: +420226201570
Re: [COOT] CHO H-bonds.
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 Dear Jan, when you input the wrong wavelength during data processing, you rescale the Ewald sphere construction, i.e. it will result in a different unit cell and crystal to detector distance. I am not sure what you mean by 'which no program from integration to structure solution did not recognised'. How should the downstream software detect the human mistake? The wavelength you enter should of course be double checked. Nowadays this seldomly occurs because the experimental settings are usually transferred automatically and reliably between the beamline software and the processing tools. Cheers, Tim On 07/02/2015 09:27 AM, Jan Stransky wrote: About the unit cell shrinkage: I have heard a lecture (I believe it was Andrea Thorn form G. Sheldrick group on one of CCP4 meetings), where she mentioned a mistake in X-ray wavelenght input (swaped 3rd and 4th digit), which no program from integration to structure solution did not recognised. It appeared in structure refinement in wrong bond distances. Worth of checking. Jan On 07/01/2015 10:58 AM, Ian Tickle wrote: Paul, as an aside to this, would it be possible to have it so that if the relevant boxes in the Environment Distances menu are checked then H-bonds and.or bumps are automatically re-calculated on re-centering? Currently it's necessary to check off on one of the boxes every time the view is re-centered. If I forget to do that it's very easy to miss bumps (and I don't have a great deal of faith in MP's concept of what constitutes a bump)! Cheers -- Ian On 30 June 2015 at 14:33, Ian Tickle ianj...@gmail.com mailto:ianj...@gmail.com wrote: Hello All I guess this is really a question about MolProbity (and possibly about autoBuster) but I assume that most Coot users will be using the MolProbity validation tools. I am in the process of depositing 4 structures of the same protein (different ligands) and I noticed that MP seems to be reporting an unusually large number of bumps in both the small overlap and bad overlap classes. In each case the resolution is 2 Ang., the structures have been refined by autoBuster and the density seems to be unequivocal, see e.g.: https://drive.google.com/file/d/0B4H4H-DyO60-SEQwS1k5S3RIVG8/view?usp=sharing A lot of the bumps are main-chain CHalpha to main-chain carbonyl O H-bonds, but there are also some CH...O side-chain H-bonds, again with clear density. The C...O distances are in the range 3.0 to 3.2 Ang., so too short for a vdW contact. The H...O distances are ~ 2.2 Ang. which is definitely shorter than the sum of the vdW radii ( H: 1.2 + O: 1.5 = 2.7). I found this survey of CH...O H-bonds but it's restricted to CH...O bonds at the end of helices and I see them mostly in sheets. http://www.mrc-lmb.cam.ac.uk/genomes/madanm/pdfs/chapter1.pdf This reports 11 examples (i.e. H-bonds, not structures) in the 3.0-3.2 range for the whole of the PDB (admittedly as it was in 2001 when the article appears to have been written). I have about the same number in one structure! One possibility I considered was that the unit cells had all somehow 'shrunk'. This can be tested with WhatCheck: however it only reports a very small shrinkage which translates to an error of ~ 0.02 Ang. in a 3 Ang. distance, which is nowhere near enough to explain a discrepancy of 0.5 Ang. So I guess my question is has anyone else noticed this in their MP dot-plots; also does anyone know what criteria does MP uses for testing bumps and specifically what value is it using for CH...O H-bonds? And of course I'd like to know whether this will affect my percentile ranking in the clashscore from the PDB validation! Cheers -- Ian - -- - -- Dr Tim Gruene Institut fuer anorganische Chemie Tammannstr. 4 D-37077 Goettingen phone: +49 (0)551 39 22149 GPG Key ID = A46BEE1A -BEGIN PGP SIGNATURE- Version: GnuPG v1 iD8DBQFVlPeFUxlJ7aRr7hoRAuQXAJ4gYxRxL/BELbrDGpAxUXbUErBPgwCghcsv nEVcp0relCBO/HVt9poCnYE= =V1FJ -END PGP SIGNATURE-
[COOT] Coot only warns of cis peps on first round of real space refinement?
Hi all, I've noticed that in the traffic light refinement validation dialogue that pops up after real space refinement, Coot warns the user when Cis peptides are introduced (Which is a very good thing!), but it does not warn the user if a cis peptide is already present in the residue range - is this the desired behaviour? I would suggest that Coot should *always* warn the user if there is a cis peptide in the refined zone, because chances are it is a mistake, and if it is not then the user can always ignore the warning. This would help when correcting a starting model with poor geometry ( I know a list of cis peps can be found in the extensions tab, but having both would be even better). Here is a screen recording illustrating the issue: https://www.dropbox.com/s/7ul1u0a6fdkl6vp/coot_cis_pep.mov?dl=0 Cheers, Oliver.
Re: [COOT] CHO H-bonds.
Hi Jan Thanks for the suggestion. I checked AFAICS the wavelength is correct. I think in any case the difference you are talking about would be far too small to explain the effect I'm seeing (the error in the wavelength is ~ 0.01% whereas I'm seeing deviations from the ideal vdW contact of up to 15%!). My conclusion is that this is not due to any kind of error, it is a real effect. CH...O bonds are well-documented (thanks to Patrick Loll and Scott Horowitz for useful references), even though it appears that MP (and PDB validation) takes absolutely no account of them! Cheers -- Ian On 2 July 2015 at 08:27, Jan Stransky stran...@ibt.cas.cz wrote: About the unit cell shrinkage: I have heard a lecture (I believe it was Andrea Thorn form G. Sheldrick group on one of CCP4 meetings), where she mentioned a mistake in X-ray wavelenght input (swaped 3rd and 4th digit), which no program from integration to structure solution did not recognised. It appeared in structure refinement in wrong bond distances. Worth of checking. Jan On 07/01/2015 10:58 AM, Ian Tickle wrote: Paul, as an aside to this, would it be possible to have it so that if the relevant boxes in the Environment Distances menu are checked then H-bonds and.or bumps are automatically re-calculated on re-centering? Currently it's necessary to check off on one of the boxes every time the view is re-centered. If I forget to do that it's very easy to miss bumps (and I don't have a great deal of faith in MP's concept of what constitutes a bump)! Cheers -- Ian On 30 June 2015 at 14:33, Ian Tickle ianj...@gmail.com wrote: Hello All I guess this is really a question about MolProbity (and possibly about autoBuster) but I assume that most Coot users will be using the MolProbity validation tools. I am in the process of depositing 4 structures of the same protein (different ligands) and I noticed that MP seems to be reporting an unusually large number of bumps in both the small overlap and bad overlap classes. In each case the resolution is 2 Ang., the structures have been refined by autoBuster and the density seems to be unequivocal, see e.g.: https://drive.google.com/file/d/0B4H4H-DyO60-SEQwS1k5S3RIVG8/view?usp=sharing A lot of the bumps are main-chain CHalpha to main-chain carbonyl O H-bonds, but there are also some CH...O side-chain H-bonds, again with clear density. The C...O distances are in the range 3.0 to 3.2 Ang., so too short for a vdW contact. The H...O distances are ~ 2.2 Ang. which is definitely shorter than the sum of the vdW radii ( H: 1.2 + O: 1.5 = 2.7). I found this survey of CH...O H-bonds but it's restricted to CH...O bonds at the end of helices and I see them mostly in sheets. http://www.mrc-lmb.cam.ac.uk/genomes/madanm/pdfs/chapter1.pdf This reports 11 examples (i.e. H-bonds, not structures) in the 3.0-3.2 range for the whole of the PDB (admittedly as it was in 2001 when the article appears to have been written). I have about the same number in one structure! One possibility I considered was that the unit cells had all somehow 'shrunk'. This can be tested with WhatCheck: however it only reports a very small shrinkage which translates to an error of ~ 0.02 Ang. in a 3 Ang. distance, which is nowhere near enough to explain a discrepancy of 0.5 Ang. So I guess my question is has anyone else noticed this in their MP dot-plots; also does anyone know what criteria does MP uses for testing bumps and specifically what value is it using for CH...O H-bonds? And of course I'd like to know whether this will affect my percentile ranking in the clashscore from the PDB validation! Cheers -- Ian -- Jan Stransky, PhD student Institute of Biotechnology, CAS Laboratory of structure and function of biomolecules Nad Safinou II 366 Vestec Czech Republic Tel.: +420226201570