Dear Dale,
Thank you for this clarification. I guess my suggestion was more based on practical experience. By taking out the waters, one already eliminates a large fraction of the atoms who have stopped diffracting to use your terminology.
This leaves me with the following question: if, as you suggest, the "calculated" (via Fcalc) Wilson plot almost by default matches the observed Wilson plot, why do people insist that one compares the Wilson plot to the average B?
Best regards,
Herman
-----Original Message-----
From: CCP4 bulletin board [mailto:[log in to unmask]] On Behalf Of Dale Tronrud
Sent: Friday, October 09, 2009 3:05 PM
To: [log in to unmask]
Subject: Re: [ccp4bb] temperature factors, mosaicity and wilson plot
The Wilson B and the average B are calculated differently so there is no reason to expect that they will be equal. The derivation of the Wilson B assumes that all B's in the model are equal. When there really is a spread in the actual values in the PDB file the derivation isn't really valid. Usually the average B will be larger than the Wilson B because the atoms with very large B values will have stopped diffracting at lower resolution than the shells used to calculate the Wilson B.
If you want to compare you model to your Wilson B you should calculate Fcalc from your PDB and calculate a Wilson B from those. The numbers will almost certainly match. The refinement program has to be pretty screwed up to cause the overall B of the model not to match the Fobs.
Dale Tronrud
[log in to unmask] wrote:
> Dear Dominik,
>
> I do not see any problems with a mosaicity of 0.48°. Also the Wilson B-factor of 30.34 of your data is ok, indicating no problems with the processing. Your problem is that the average REFINED b-factor is much higher than your "observed" average b. The first thing I would do is to calculate the average b-factor for your protein only. Adding a lot of high b-factor water molecules can increase your average b-factor significantly. If the average- and wilson B are then in the same ballpark, I would not worry too much and rahter have a look at my waters to see if some of them could be removed. If the descrepancy is still there for the protein only, I would look at the way the b-factors are refined (individual, grouped, TLS, weights, restraints etc.) to see if there is a problem somewhere.
>
> Best regards,
> Herman
>
> -----Original Message-----
> From: CCP4 bulletin board [mailto:[log in to unmask]] On Behalf Of
> Dominik Possner
> Sent: Thursday, October 08, 2009 5:59 PM
> To: [log in to unmask]
> Subject: [ccp4bb] temperature factors, mosaicity and wilson plot
>
> Hi,
>
> I recently solved the structure of a mutant protein that contains a single amino acid substitution.
>
> Here are some data about the mutant:
>
> resolution: 2.65 A
> completeness: 99.9 % (100.0 %)
> Rwork/free: 22.3/27.8 %
> Rsym 15.6 (47.7) %
> avg. mosaicity: 0.48°
>
> now whats confusing me:
> Wilson b-factor: 30.34 A^2
> Overall mean b-factor: 55.92 A^2
>
> My questions are the following:
>
> Where could the large difference of the b-factors come from? As noted, the crystal shows a mosaicity of avg. 0.48°.
>
> Is there a way I could evaluate whether the high b-factors are completely explained by the high mosaicity? What other factors could explain the high b-factors? Does the b-factor correlate that clearly with the resolution?
>
> I used XDS for data processing and REFMAC for refinement. Does either of these programs consider the mosaicity so that i could exclude that the b-factors origin from the mosaicity?
>
>
> Thanks,
> Dominik
