Hi Bernhard, You touch a valid point. You get different clash scores in MP depending on how you add hydrogens when you run it (unless this was solved recently). So you can essentially cheat the program by choosing a different H addition function. I don't get why this option exists at all. If you add hydrogens to your model you should do it in such a way that your validation algorithm gives a reliable answer.
For what it is worth, I think hydrogens should be modelled where they are, not where we observe them. That means for X-ray we need a correction along the the bond, for neutron diffraction we don't, and for electron diffraction, well that gets a bit more complicated. Cheers, Robbie Sent from my Windows 10 phone Van: Bernhard Rupp<mailto:[email protected]> Verzonden: zaterdag 29 april 2017 21:18 Aan: [email protected]<mailto:[email protected]> Onderwerp: Re: [ccp4bb] CH-bond length discrepancies Thanks to all respondents. It seems the source of the discrepancy is that some programs use nuclear distances and some use electron density peak distances. Now, does it matter, for example, using nuclear distances for placing riding Hs in the VdW terms, (i.e., a model for dipole interactions which are largely electronically dominated) and then validate against X-ray (i.e. electronic) distances? If this use of different distance sets were consistent (I am not even sure of that), then we would probably introduce systematic bias. The point is probably academic in practical terms but somehow it would be interesting to understand how using which distances when is justified… Best, BR From: Zhijie Li [mailto:[email protected]] Sent: Friday, April 28, 2017 3:42 PM To: [email protected] Subject: Re: [ccp4bb] CH-bond length discrepancies Hi Bernhard, I believe this is the data you are interested: http://journals.iucr.org/b/issues/2010/03/00/so5037/so5037.pdf Zhijie On 28/04/2017 1:33 PM, Bernhard Rupp wrote: Dear Fellows of the Bond, when validating a QM refined homology model with Molprobity, I noticed various 8 sigma deviations in the carbon-hydrogen bond distances. Out of curiosity, I then used refmac to calculate riding Hs for the same model, and at least in one instance (N-H backbone) there are significant differences between Molprobity and Refmac H bond distances (differences to the QM distances in other instances I find interesting, but less relevant for us). The riding H vs Molprobity presumably should be consistent, because if we use them in VDW restraints but they differ from the validation target, systematic bias will occur. I have no feel how significant that effect might be – maybe someone more erudite can comment. Examples distance MP REF QM backbone N-H 0.86 1.01 1.00 phenyl C-H 0.93 0.93 1.09 Best, BR PS: If someone has accurate experimental values for CH distances I’d appreciate a link. No access to CSD. ------------------------------------------------------ Bernhard Rupp Crystallographiae Vindicis Militum Ordo http://www.hofkristallamt.org/ [email protected]<mailto:[email protected]> +1 925 209 7429 +43 767 571 0536 ------------------------------------------------------ Many plausible ideas vanish at the presence of thought ------------------------------------------------------
