Dont forget the confusing interaction between recorded B factors in a
PDB file and TLS parameters. This Q is still I think unresolved but you
need to be careful!
Eleanor
Edward A. Berry wrote:
Interesting that the correlation between B-factor and
resolution didn't show up in the QDB as recently as 2001.
Or am I missing something?
From what i understand this includes structures with
resolution from 1.5 to 3.5 A.
On Jan 4 2001 Gerard "CD" Kleywegt wrote:
well, i had a quick look at the data stored in QDB
(gjk, acta cryst d52, 842-857) which shows that
for 435 structures the corr coeff between resolution
and average b is only 0.06, i.e. insignificant
the only non-trivial correlate (using a 0.2 cut-off)
is the percentage of secondary structure (makes
sort of sense) with cc=0.20
in my other large-scale test, mentioned a couple
of weeks ago, i found that essentially all temp-
factor-related staistics are "incorrectly"
correlated with measures of model accuracy
(e.g., higher average b tends to be accompanied
by higher accuracy !). average b is very strongly
correlated with completeness on the other hand.
i suspect that problems with data and/or restraints
(rather than physics) are a major determinant
of the temp-factors we calculate for our models ...
--dvd
On Thu, 4 Jan 2001, Yu Wai Chen wrote:
Does any one know if there is any correlation between the overall
B-factor of a structure in relation to its resolution? Are there any
publications on this topic?
Also is there any correlation between the extent of disorder in a
structure and the R-factor/Rfree?
Pavel Afonine wrote:
Hi,
all the errors go into B and so you can get decent R with wrong
structure. Glycosylated proteins have a large component totally
disordered - do you see any sugars?
with B~133 you uj is 3.7A which means that atom is all over the
place and meaningless
As a reviewer I would certainly question the interpretation of
such structure.
If the data collection was at 100K (or any cryo condition) one
expects B~20 for a good structure
I wouldn't interpret it that literally. The distribution of mean
B-factors computed for structures in PDB at resolutions between 2.8
and 3..2A is:
Mean B-factor value Number of
structures in PDB
1.000 - 21.000 : 137
21.000 - 41.000 : 430
41.000 - 61.000 : 612
61.000 - 81.000 : 390
81.000 - 101.000 : 172
101.000 - 121.000 : 36
121.000 - 141.000 : 12
141.000 - 161.000 : 8
161.000 - 181.000 : 0
181.000 - 201.000 : 1
There is good amount of models with mean B-factors well higher than
70. I doubt that PDB would accept a structure where atoms are "all
over the place" -:) And I wouldn't claim that those models are all
bad simply because they don't have B-factors~20.
There is a number of publications that discuss this and show similar
histograms, so there is no point to repeat it.
So, Jiamu, *if* the high B-factors is the only issue with this part
of your structure, then make sure that the domain in question is
properly modeled, and keep the above histogram just in case should
you run into a picky reviewer -:)
Pavel.
PS>
Same histogram as above, but computed for all models in resolution
range from 3 to 4A:
Mean B-factor value Number of
structures in PDB
0.000 - 22.720 : 90
22.720 - 45.440 : 250
45.440 - 68.160 : 295
68.160 - 90.880 : 271
90.880 - 113.600 : 131
113.600 - 136.320 : 73
136.320 - 159.040 : 32
159.040 - 181.760 : 13
181.760 - 204.480 : 8
204.480 - 227.200 : 5
and finally, for high resolution models in 0.0 to 1.0A:
Mean B-factor value Number of
structures in PDB
1.800 - 4.190 : 3
4.190 - 6.580 : 4
6.580 - 8.970 : 18
8.970 - 11.360 : 41
11.360 - 13.750 : 50
13.750 - 16.140 : 34
16.140 - 18.530 : 9
18.530 - 20.920 : 3
20.920 - 23.310 : 3
23.310 - 25.700 : 5
