B average lower than Wilson

From: anna anna
Date: 5 March 2012 15:16
Dear all, I hope you can help me!
I solved a structure at 3.0 A res and everything is quite satisfactory, my only concern is the discrepancy between
the B factor from Wilson plot (58.8 A^2) and the average B (36.9). It seems strange, usually the opposite occurs...
What does it mean? Which may be the reason? What can I check?
Some details:
Rmerge: 15.0 (51.8)
I/σ(I): 7.30 (1.84)
5 monomers/AU
R/Rfree: 19.7/23.3
...ask for other infoes if you need.

Any advice/suggestion is welcome!! 

----------
From: Tim Gruene


Dear     ,

the Wilson plot is only meaningful with data better than 4A. Since your
data extends to only 3A (and according to your I/sig(I) stats you pushed
that limit a little...), you may have too little data for a good fit.

That's only one possibility. Your data may have ice rings resulting in
poor intensity estimates, are the atoms may indeed have low B-values and
the reason for limiting the resolution may have a different source than
usual,...

Tim

- --
-

----------
From: Pius Padayatti
Hi,
Please see a similar thread at ccp4bb earlier at the following link

http://www.proteincrystallography.org/ccp4bb/message16011.html

Padayatti

----------
From: anna anna
Dear all, thanks for help!
I'm using Refmac without TLS.
Tim's explanation is pretty convincing to me. My data have no ice rings compromising the extimation of I,  but  the map, is more defined then I expect at 3 A resolution, so a "low" average B seems consistent. Which can be other causes for limited resolution?
Anna 

----------
From: Eleanor Dodson
Not an answer, but if you look at the Wilson plot after scala or truncate, you can see how good the estimate is - at 3A those can be pretty awful Eleanot

--