Hi Pavel glad you write me. I was hoping you would read my post.
- Yes, protons are added, both on the protein as well as on the molecule - I initially only refined protein and ligand anisotropically, now Im running a refinement with all atoms anisotrp except Hs. This would then also be the same as shelxl is doing. - Alternate conformations are modeled, also on the ligand. There are plenty, sure, but I think I got most of them. - I already used Water update during refine, there are some NO3s in the structure. I got them in. There is a second ligand somewhere as artifact. its density is not well defined, so I hope to get that in once the map clears up more. - I let phenix.refine optimize adp and chemisty weights, but as Petri suggested, Im manually increasing the scale factors to match the ones from shelxl (just to compare them properly). Im aiming for an rsmd of 0.02-0.03A like Petri suggested and keep an eye on how tight the structure is refined in shelxl. About the Rfact and the gap. Yes, thats what I was expecting. I hope if I add more anisotropic B fact, the Rfacts should go down to at least what shelxl yielded. thank you all again for the massive feedback, ideas and help. Dr. Matthias Barone AG Kuehne, Rational Drug Design Leibniz-Forschungsinstitut für Molekulare Pharmakologie (FMP) Robert-Rössle-Strasse 10 13125 Berlin Germany Phone: +49 (0)30 94793-284 ________________________________ From: Pavel Afonine <[email protected]> Sent: Monday, February 3, 2020 7:14:25 PM To: Barone, Matthias Cc: [email protected] Subject: Re: [ccp4bb] refinement of 0.73A data in shelxl Hi Matthias, did you use correct model parameterization and optimal refinement strategy for the resolution? Such as: - Add H atoms; - Refine all but H atoms with anisotropic ADPs; - Model alternative conformations (that one'd expect many at this resolution); - Add solvent (water, crystallization cocktail components if you see any); - Relax restraints on geometry and ADPs; .... long list! If not, then what you have in terms of R factors is more or less what I'd expect. In the absence of obvious data pathologies, I'd expect Rwork/Rfree in 10-15% range, and the Rfree-Rwork gap around 1-2% or less. Since you mentioned Phenix refinement, I am happy to help you with details etc off-list. Pavel On Mon, Feb 3, 2020 at 3:08 AM Barone, Matthias <[email protected]<mailto:[email protected]>> wrote: Dear ccp4 community Im having some problems solving a 0.73A structure. Spacegroup seems to be correct, data are not twinned, 95.5% overall completeness, ISa 25.6. Outer shell CC1/2 24% and 90.4% complete. The model is nearly fully built, there is no remaining unmodelled areas. However, Rfact is stuck 27% in phenix, with a very distinct artifact in the electron map (see phenix.jpg). You can see difference density on various well defined sidechain atoms. Notably, they seem to follow a pattern: Nearly all Val CG have difference signal, as well as many backbone NH. Hence, I suspected that it might be a problem with the SF, since we recorded the DS at 0.86A. Hence I gave shelxl a shot: I used the refined model from phenix, converted it via pdb2ins and pasted the restraints created by prodrg. The shelxl hkl was produced by xdsconv, using the freeR flagging of the mtz used by phenix (no merge, friedel false). Interestingly, shelxl can bring Rfree down to 16% and almost all of the diff-density artifacts seen before are gone (shelxl_noSFAC-CL.jpg). Except one: the inhibitor contains a chlorinated phenylring (pdb ligand 2L5) which now shows massive difference density for Cl. I therefore suggested that I might deal with a wrong SF for Cl. Funny enough, pdb2ins does not produce a DISP line for Cl if converting the pdb that contains the inhibitor. Hence, I used pdb2ins and the pdb from PRODRG to produce SFAC for the inhibitor Cloride. I then pasted this line DISP $CL 0.18845 0.21747 1035.16450 into the .res file and updated the UNIT line. Shelxl runs through, and the density looks ok on the Chloride now. However Rfree is back up at 24% and the artifacts seen by phenix.refine are back (shelxl_SFAC-CL.jpg): now, very distincitvly, backbone carbonyls and NHs show difference density. Am I right in my assumption, that the SFAC of Cloride is not properly calculated at the given wavelenght? And if so, how do I guess it correctly? Thank you very much for your help! Best, matthias Dr. Matthias Barone AG Kuehne, Rational Drug Design Leibniz-Forschungsinstitut für Molekulare Pharmakologie (FMP) Robert-Rössle-Strasse 10 13125 Berlin Germany Phone: +49 (0)30 94793-284 ________________________________ To unsubscribe from the CCP4BB list, click the following link: https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCP4BB&A=1 ######################################################################## To unsubscribe from the CCP4BB list, click the following link: https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCP4BB&A=1
