Re: [ccp4bb]: WHATCHECK Deformation Matrix - is it advisable to c orrect the cell ?

[Dirk Kostrewa]

Hi Jorge, I stumbled across the same deformation matrix problem reported by WHATCHECK a couple of years ago and initiated a little discussion on the CCP4BB, at that time assuming a bug in the program. However, from discussions with the author of WHATCHECK, Gerd Vriend, it turned out that apparently this deformation matrix resulted from slightly different implementations of the same (!) Engh & Huber parameters in WHATCHECK and, by that time, in XPLOR/CNS. I can't remember anymore, which of the bond lengths came out slightly different. But let's assume, that the C=O bonds on average come out slightly longer in your refinement program compared to the library used by WHATCHECK (there are some doubts about the correct length of the C=O bond in E&H). If you don't have a single (anti-)parallel helix bundle or beta-sheet as the only structural feature of your protein in the unit cell, then the directions of the C=O-vectors should be more or less equally distributed with respect to your coordinate system, meaning that also all components of these vectors along the unit cell axes should occur with about equal frequencies. A systematic comparison of the on average too long C=O-bond lenghths with the WHATCHECK library value would then suggest, that your unit cell dimensions should be decreased by a few percent, so that after orthogonalization the refined C=O bonds come out with the "correct" slightly shorter average length (I hope, it's not the other way around ;-) ). I can't tell you exactly how WHATCHECK does its analysis, because the web-site is currently not reachable. As long as the data processing was done with great care, personally, I would trust the refined unit cell parameters more than the "deformation matrix" analysis by WHATCHECK. Regarding your question about cryo-temperature bond lenghts, I think, it would be time to do a new analysis of ideal bond lenghts of now many more very high resolution protein structures whose crystals were measured at cryo-temperature to complement the Engh & Huber parameters. Best regards, Dirk. Jorge Iulek wrote: Dear Friends, If possible, I would like to have some more explanationon this question. Hi Alex, as far as I understand, the WHATCHECK deformation matrix tries to figure out any systematic deviations from ideal stereochemistry that could be the result of systematically wrong cell dimensions. In practice, it As systematically here you mean in all 3 cell parameters ? What if (:-) ) you had one overestimated and one underestimated ? I think WHATCHECK must somehow separate bond lenghts (eg) in each of the 3 components and from this it takes its conclusions about each cell parameter. Am I right ? usually detects systematic deviations between the built-in stereochemistry library What does built-in mean here ? Built-in in WHATCHECK ? Should it not be the Engh-Huber parameters, like for instance the one in refmac5 ? and the one that was used during refinement. Yet one other question here is how should we expect a cell dimension to change under cryocooling ... When you modify the cell constants as suggested in the WHATCHECK output and do a re-refinement, the next output will have similar complaints, and so on. If you took care about proper data processing, the cell constants should be usually well determined. In such a case, at least in my experience, you can safely ignore the deformation matrix warnings in WHATCHECK. It is good to listen to Dirk's experience. Is it common sense that we should, unless serious erros are around (eg, what about smeared spots in the image, etc..), ignore WHATCHECK recommendation and go for the refined cell in the image processing program ? Jorge Best regards, Dirk. Alex Gutteridge wrote: *** For details on how to be removed from this list visit the *** *** CCP4 home page http://www.ccp4.ac.uk *** Hi, I'm using WHATCHECK (http://swift.cmbi.ru.nl/gv/pdbreport/) to scan some crystal structures for unit cell scaling problems. I'm doing this because the analysis I want to do on these structures will be sensitive to these kind of errors. At this point I should mention that I'm not a crystallographer by trade, so forgive me if this is an obvious/stupid question. WHATCHECK gives a 'deformation matrix' as part of it's output in this section. Am I right in thinking that if the matrix looks like this (from PDB code 2DHB): 0.987651 -0.000042 -0.000855 -0.000042 0.985582 -0.000365 -0.000855 -0.000365 0.987811 Then it is (essentially saying) the structure as given in the PDB file is stretched about 1-2% (scale factor of 0.99-0.98) along each axis? And so any measurement of length/distance in the structure will be 1-2% longer than it should be? Also does anyone know of a program that can apply this matrix to a PDB file to get the correctly scaled structure. I've found one paper that implies they did this, but they don't really say how. Thanks in advance! Alex Gutteridge EBI Wellcome Trust Genome Campus Hinxton Cambs CB10 1SD UK Tel: 01223 492546 Email: [EMAIL PROTECTED] -- **************************************** Dirk Kostrewa Paul Scherrer Institut Life Sciences, OFLC/110 CH-5232 Villigen PSI, Switzerland Phone: +41-56-310-4722 Fax: +41-56-310-5288 E-mail: [EMAIL PROTECTED] http://sb.web.psi.ch **************************************** -- **************************************** Dirk Kostrewa Paul Scherrer Institut Life Sciences, OFLC/110 CH-5232 Villigen PSI, Switzerland Phone: +41-56-310-4722 Fax: +41-56-310-5288 E-mail: [EMAIL PROTECTED] http://sb.web.psi.ch ****************************************