4 Feb
2008
4 Feb
'08
8:26 p.m.
Hi Pavel, thanks for the reply: 1.The maps look great for the residues were some of the B-factors are 0. Visually there are no anomalies in the maps. For residues with a large B-factor distribution (10-20 for mainchain, >120 for carboxylic oxigens), i.e. surface GLU and ASP, they look as expected. Very well defined for mainchain, no or very faint density for the last side chain atoms. 2. Did that. As I mentioned in the mail, I started out from a refined model, set all B-factors to the Wilson B and then refined B-factors for 10 macrocycles using wxu optimization. ==> Same result. 3. OK, I'll play around with that. Any idea what target I should use? Minimize R-free or just push all Bs above 0? Does anyone know a publication which covers theoretical B-factor distributions? Cheers, Carsten > -----Original Message----- > From: [email protected] > [mailto:[email protected]]On Behalf Of Pavel Afonine > Sent: Monday, February 04, 2008 2:08 PM > To: PHENIX user mailing list > Subject: Re: [phenixbb] Question abount B-factor refinement - > B-factors > = 0 > > > Hi Cartsen, > > I agree that zero B-factors is a warning sign. I would try do > the following: > > 1) Look at the maps to check how they look like for the > problem fragments. > > 2) You can try to run "optimize_wxu" to get the "best" weight between > Xray target and B-factor restraints. Note that in this case > "best" means > lowest Rfree which does not guarantee the distribution of > B-factors that > you will like. > > 3) Manually tighten B-factor restraints until you get the > distribution > of B-factors that you think is reasonable.You can do this by either > decreasing the default value of "wxu_scale" or increasing the > value of > "wu". Just to remind that for B-factors refinement the > overall target is > defined as > > Etotal = Exray * wxu * wxu_scale + Eadp * wu > > where Exray is Xray target, wxu - automatically determined weight, > wxu_scale - adjustment scale (that user can adjust), wu - same as > wxu_scale applied to Edp (B-factor restraints). > > Without playing with the refinement myself I can't say more. Once the > structure is published, I would be very interested to play > with it since > it seems like a good case to use it for further improvement of > phenix.refine. > > Pavel. > > > > > > Initially, I noticed that the B-factors within a residue > have quite a large spread, when one runs a temperature factor > variance analysis as implemented in coot (Yes I read Pavel's > response about this ...). This prompted me to actually look > at the B-factors in the PDB file. I noticed that actually for > quite a few residues one or more of the B-factors refines to > 0.00 For instance one of the less extreme examples: > > > > ATOM 118 N HIS A 21 -28.768 35.430 2.317 > 1.00 4.73 N > > ATOM 119 CA HIS A 21 -27.357 35.565 2.625 > 1.00 0.00 C > > ATOM 120 C HIS A 21 -27.124 35.408 4.116 > 1.00 2.25 C > > ATOM 121 O HIS A 21 -28.072 35.360 4.902 > 1.00 7.73 O > > ATOM 122 CB HIS A 21 -26.853 36.940 2.186 > 1.00 8.10 C > > ATOM 123 CG HIS A 21 -27.610 38.078 2.799 > 1.00 11.18 C > > ATOM 124 ND1 HIS A 21 -28.733 38.624 2.216 > 1.00 17.18 N > > ATOM 125 CD2 HIS A 21 -27.418 38.759 3.955 > 1.00 13.28 C > > ATOM 126 CE1 HIS A 21 -29.194 39.600 2.979 > 1.00 25.70 C > > ATOM 127 NE2 HIS A 21 -28.414 39.701 4.041 > 1.00 12.79 N > > > > > > As background the data extends to 1.8A, with an I/I(s) of > ~15 and ~2.5 in the highest shell, CHI^2 are all around 1+/ > 0.1. The Wilson B is 11.3A^2 and R/Rfree 16.8/19.8%. This is > homesource date, at the synchrotron I could probably push the > resolution further out to better than 1.6A. No hydrogens were > added, nor was TLS used. > > > > In order to get around the problem I initally reset all > B-values to the Wilson-B, enabled wxc refinement (latest > CCI-APPs) and refined ADPs only for 10 macrocycles. After a > couple of cycles the B-factor distribution settled into the > following, which looks fairly reasonable to me, except for > the min values of 0.00: > > > > |-ADP statistics (Wilson B = > 11.289)------------------------------------------| > > | Atom | Number of | Isotropic or equivalent| > Anisotropy lmin/max | > > | type |iso aniso | min max mean | min > max mean | > > | - - - - |- - - - - - -| - - - - - - - - - - - -| - - - > - - - - - - - | > > | all : 2659 0 0.00 129.99 13.58 None > None None | > > | all(noH): 2659 0 0.00 129.99 13.58 None > None None | > > | Sol. : 390 0 0.35 56.99 27.75 None > None None | > > | Mac. : 2269 0 0.00 129.99 11.14 None > None None | > > | Hyd. : 0 0 None None None None > None None | > > | - - - - - - - - - - - - - - - - - - - - - - - - - - - > - - - - - - - | > > | Distribution of isotropic (or equivalent) ADP for > non-H atoms: | > > | Bin# value range #atoms | Bin# value > range #atoms | > > | 0: 0.000 - 12.999: 1778 | 5: 64.993 - > 77.991: 17 | > > | 1: 12.999 - 25.997: 492 | 6: 77.991 - > 90.990: 9 | > > | 2: 25.997 - 38.996: 210 | 7: 90.990 - > 103.988: 4 | > > | 3: 38.996 - 51.994: 124 | 8: 103.988 - > 116.987: 2 | > > | 4: 51.994 - 64.993: 19 | 9: 116.987 - > 129.985: 4 | > > | =>continue=> > | > > > |------------------------------------------------------------- > ----------------| > > > > > > B-factors of 0 indicate to me that there is a problem, not > sure how to tackle this in phenix at this point. Could there > be a problem with the bulk-solvent scaling? Should I restrict > the minimum B-factor to 1.0 (ugggh)? Or should I tighten the > B-factor restraints? > > > > Cheers, > > > > Carsten > > > > > > > > > > > > _______________________________________________ > > phenixbb mailing list > > [email protected] > > http://www.phenix-online.org/mailman/listinfo/phenixbb > > > _______________________________________________ > phenixbb mailing list > [email protected] > http://www.phenix-online.org/mailman/listinfo/phenixbb > >