[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: aniso scaling
*** For details on how to be removed from this list visit the ***
*** CCP4 home page http://www.dl.ac.uk/CCP/CCP4/main.html ***
There's nothing wrong with high B-factors in a structure. In fact, if they're
not high when your resolution is not brilliant, Be Afraid. Your model has to
describe the data, and B-factors are part of the model, indicating
uncertainty. If your data implies high B-factors, then that's what your model
should have. In fact, if you try to force the B-factors to remain low by some
artificial means, you're creating an inaccurate model. The B-factor is no
indication of the accuracy of a model.
Of course, if you start off refinement with B-factors far from the right
values, they're going to have trouble converging properly, so resetting them to
a more realistic value (from the Wilson plot, for instance) before refining
them would possibly sort out your funny B distribution. But even that need not
be a problem: if it makes sense in terms of the structure (maybe a buried side
chain coming from an exposed main chain?) then that's what it should be.
Phraenquex
> Dear All,
>
> I am refining a structure at 2.3 A resolution. I have a tetramer in the
> asymmetric unit. I am using STRICT NCS and group B's (2 per residue: main
> chain atoms and side chains atoms).
> I always start a refinement using XPLOR and I finish it with REFMAC because
> I find it easier to manipulate alternate conformations (that won't happen in
> this case!!) but also because the anisotropic scaling seems to work better.
> Is it?
>
> The scaling in XPLOR gives:
>
> SCALEF: k= 0.170 B11= -1.221 B22= 18.857 B33= 7.993 B12= 0.000
> B13= -10.499 B23= 0.000
>
> It sure looks like it needs some aniso scaling.
>
> That makes my R and Rfree drop almost 4% each after bulksolvent correction,
> powell minimization and some B refinement. However, in XPLOR I noticed that
> after the anisotropic scaling my B's go up and up and up (before average B
> is 28 and after scaling, it's 40!!!!! And it continues to go up after each
> cycle). The problem is I don't know what is going on with those B and I
> don't have any confidence in them. They also don't make any sense. Some of
> the main chain atoms will have B's higher than the side chain atoms in the
> same residue.
> This happens not only for this structure but for all of the one I work on.
> It is also obvious when you retrieve some of the pdb files in the Data Bank.
> Am I the only one worried about it?
>
> Can someone shed some light on that matter please?
>
> Here are statistics from Moleman for before and after some refinement.
>
> Before:
>
> powell16c R = 0.306181 FREE-R = 0.320428 15-2.3 A
> with 2.0 sigma cutoff
> ====================================================================
> Chain name Atoms Bave Bsdv Bmin Bmax
> <->A1 942 28.61 10.36 13.20 73.83
> <->A2 804 32.89 10.33 14.71 65.89
> <->AL 23 32.05 2.54 27.81 38.12
> <->AM 1 35.18 1237.63 35.18 35.18
> <->AW 15 34.38 0.00 34.38 34.38
> ====================================================================
>
> After:
>
> powell17c R = 0.26149 FREE-R = 0.283608 aniso scaling 15-2.3
> A with 2.0 sigma cutoff
> ====================================================================
> Chain name Atoms Bave Bsdv Bmin Bmax
> <->A1 942 40.91 11.09 18.80 85.51
> <->A2 804 45.33 10.73 22.57 82.86
> <->AL 23 39.15 5.76 30.00 50.88
> <->AM 1 46.41 2153.89 46.41 46.41
> <->AW 15 46.80 0.01 46.80 46.80
> Nr of chains encountered : ( 5)
> ====================================================================
>
> I used the XTALMACRO:scalef in XPLOR.
>