Jose Antonio,
I've seen similar behavior few years ago with grouped B-factor
refinement in CNS. The argument for the grouped refinement is as
follows:
This is better than individual B-factor refinement at low resolution
because you significantly reduce the number of parameters.
There are two holes in this argument.
1. The reduction in number of parameters is not that significant. On
average, you have some 8 atoms per residue, and thus 32 parameters with
individual B-factors (4 per atom). With grouped B-factors you have 26,
about 20% difference. While it sounds like a lot (hey, I just removed
2000 parameters from my 350 residue model!), it only has an effect
similar to improving resolution by ~6%, e.g. from 2.8A to 2.65A. Worthy
goal, no doubt, but will hardly lead to dramatic improvement that would
justify these crazy jumps in B-factors.
2. The total number of parameters is not the same as the effective
number of parameters. Since individual B-factors are restrained, they
don't really contribute one degree of freedom per atom. I don't know of
any reliable estimate of effective number of parameters in the
literature, but my personal guess based on several ways to estimate it
is that it may take about 5 restraints to compensate for one parameter.
B-factors are restrained by bond and by angle, thus giving 2-3
restraints per B-factor. So, every individual B-factor only contributes
half a parameter, thus ~4 parameters per residue. This is very much
comparable with two essentially unrestrained (as confirmed by their wild
variations) grouped B-factors.
Certainly, these are guesstimates. For instance, at low resolution you
apply tighter restraints and thus individual B-factors may contribute
even fewer effective degrees of freedom, perhaps even ~2 per residue.
But the general point remains that you are not necessarily better off
with grouped B-factors, even at low resolution.
By the way, I always assumed (based on B-factor behavior) that grouped
B-factor refinement is unrestrained. If it is not (in CNS and Phenix),
it is certainly not doing a good job restraining backbone-to-sidechain
and residue-to-residue variations. One thing you could do is to
calculate average B-factors per group when you refine B-factors
individually, which can be done with baverage (Structure
analysis->Temperature factor analysis in ccp4i), and then compare it to
grouped B-factors. Maybe it does vary a lot.
These B-factors look awfully high. As someone who had to deal once with
the model where <B> was about 100A^2, I feel your pain :)
Cheers,
Ed.
On Thu, 2010-01-28 at 10:34 +0100, Jose Antonio Cuesta Seijo wrote:
> Dear all,
>
> I am refining a 3.1Å structure with Phenix.refine, using two ADP
> groups per residue.
> When doing refinement in Refmac, the way to go would be to tighten the
> weights quite a bit to make up for the low resolution, resulting in
> small deviations in bond lenghts and angles from the ideal values and
> also in quite small ADP variations from atom to atom, typically in the
> range of, say, 5%.
> Now, doing the same refinement with Phenix.refine and 2 ADP groups per
> residue, the manual claims that the weights do not normally need to be
> touched. Judging by the values of R and R free (19.5% and 23.9%)
> compared to other protocols and with Refmac, it certainly does a good
> job. But the spread in ADP values in the refined model is strikingly
> high. Below is a extract from the pdb file. Note for example the jump
> from B=81.7 to B=163.8 for the main chains of ILE180 and LYS181, or
> the 163.8 to 113.5 between the main chain and the side chain of
> LYS181. Similar examples are all over the 1000+ residues in this
> structure.
> Is this normal? All global quality indicators look OK to me...
>
> Cheers,
>
> Jose Antonio Cuesta Seijo.
>
> ATOM 8338 N ARG F 178 65.398 30.884 -0.261 1.00 84.90
> N
> ATOM 8339 CA ARG F 178 66.532 31.758 -0.521 1.00 84.90
> C
> ATOM 8340 CB ARG F 178 67.576 31.628 0.583 1.00131.99
> C
> ATOM 8341 CG ARG F 178 67.044 31.967 1.952 1.00131.99
> C
> ATOM 8342 CD ARG F 178 68.084 32.695 2.778 1.00131.99
> C
> ATOM 8343 NE ARG F 178 67.464 33.433 3.873 1.00131.99
> N
> ATOM 8344 CZ ARG F 178 66.894 34.627 3.737 1.00131.99
> C
> ATOM 8345 NH1 ARG F 178 66.865 35.222 2.551 1.00131.99
> N
> ATOM 8346 NH2 ARG F 178 66.350 35.224 4.787 1.00131.99
> N
> ATOM 8347 C ARG F 178 67.152 31.347 -1.839 1.00 84.90
> C
> ATOM 8348 O ARG F 178 66.738 30.357 -2.443 1.00 84.90
> O
> ATOM 8349 N ILE F 179 68.143 32.107 -2.287 1.00 80.99
> N
> ATOM 8350 CA ILE F 179 68.954 31.666 -3.406 1.00 80.99
> C
> ATOM 8351 CB ILE F 179 69.698 32.817 -4.083 1.00 53.62
> C
> ATOM 8352 CG1 ILE F 179 68.754 33.995 -4.333 1.00 53.62
> C
> ATOM 8353 CD1 ILE F 179 68.390 34.216 -5.806 1.00 53.62
> C
> ATOM 8354 CG2 ILE F 179 70.308 32.325 -5.389 1.00 53.62
> C
> ATOM 8355 C ILE F 179 69.970 30.657 -2.884 1.00 80.99
> C
> ATOM 8356 O ILE F 179 70.677 30.914 -1.913 1.00 80.99
> O
> ATOM 8357 N ILE F 180 70.026 29.500 -3.528 1.00 81.67
> N
> ATOM 8358 CA ILE F 180 70.864 28.417 -3.061 1.00 81.67
> C
> ATOM 8359 CB ILE F 180 70.161 27.066 -3.218 1.00 71.36
> C
> ATOM 8360 CG1 ILE F 180 68.850 27.090 -2.446 1.00 71.36
> C
> ATOM 8361 CD1 ILE F 180 68.964 27.820 -1.114 1.00 71.36
> C
> ATOM 8362 CG2 ILE F 180 71.038 25.955 -2.707 1.00 71.36
> C
> ATOM 8363 C ILE F 180 72.180 28.418 -3.811 1.00 81.67
> C
> ATOM 8364 O ILE F 180 73.218 28.106 -3.240 1.00 81.67
> O
> ATOM 8365 N LYS F 181 72.137 28.768 -5.092 1.00163.76
> N
> ATOM 8366 CA LYS F 181 73.356 28.961 -5.872 1.00163.76
> C
> ATOM 8367 CB LYS F 181 74.053 27.627 -6.155 1.00113.58
> C
> ATOM 8368 CG LYS F 181 73.432 26.812 -7.280 1.00113.58
> C
> ATOM 8369 CD LYS F 181 74.307 25.615 -7.656 1.00113.58
> C
> ATOM 8370 CE LYS F 181 74.291 24.546 -6.571 1.00113.58
> C
> ATOM 8371 NZ LYS F 181 75.015 23.306 -6.984 1.00113.58
> N
> ATOM 8372 C LYS F 181 73.033 29.687 -7.172 1.00163.76
> C
> ATOM 8373 O LYS F 181 71.889 29.673 -7.629 1.00163.76
> O
> ATOM 8374 N LYS F 182 74.039 30.325 -7.760 1.00126.44
> N
> ATOM 8375 CA LYS F 182 73.849 31.088 -8.988 1.00126.44
> C
> ATOM 8376 CB LYS F 182 74.588 32.421 -8.897 1.00131.41
> C
> ATOM 8377 CG LYS F 182 74.235 33.203 -7.652 1.00131.41
> C
> ATOM 8378 CD LYS F 182 74.885 34.575 -7.634 1.00131.41
> C
> ATOM 8379 CE LYS F 182 74.466 35.367 -6.393 1.00131.41
> C
> ATOM 8380 NZ LYS F 182 75.076 36.730 -6.349 1.00131.41
> N
> ATOM 8381 C LYS F 182 74.338 30.303 -10.192 1.00126.44
> C
> ATOM 8382 O LYS F 182 75.506 29.930 -10.252 1.00126.44
> O
> ATOM 8383 N LEU F 183 73.447 30.052 -11.146 1.00161.61
> N
> ATOM 8384 CA LEU F 183 73.819 29.331 -12.360 1.00161.61
> C
> ATOM 8385 CB LEU F 183 72.591 29.005 -13.214 1.00100.55
> C
> ATOM 8386 CG LEU F 183 71.470 28.218 -12.534 1.00100.55
> C
> ATOM 8387 CD1 LEU F 183 70.602 27.516 -13.569 1.00100.55
> C
> ATOM 8388 CD2 LEU F 183 72.058 27.219 -11.558 1.00100.55
> C
> ATOM 8389 C LEU F 183 74.810 30.141 -13.183 1.00161.61
> C
> ATOM 8390 O LEU F 183 74.988 31.337 -12.943 1.00161.61
> O
> ATOM 8391 OXT LEU F 183 75.443 29.616 -14.103 1.00100.55
> O
> ______________________________________________________________________
> ***************************
> Jose Antonio Cuesta-Seijo
>
> Biophysical Chemistry Group
> Department of Chemistry
> University of Copenhagen
> Tlf: +45-35320261
> Universitetsparken 5
> DK-2100 Copenhagen, Denmark
> ***************************
--
Edwin Pozharski, PhD, Assistant Professor
University of Maryland, Baltimore
----------------------------------------------
When the Way is forgotten duty and justice appear;
Then knowledge and wisdom are born along with hypocrisy.
When harmonious relationships dissolve then respect and devotion arise;
When a nation falls to chaos then loyalty and patriotism are born.
------------------------------ / Lao Tse /
|
