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