[buster-discuss] Difference in R-factors by Buster and DCC

[BUSTER Developers]

Dear Italo,

On Wed, Apr 01, 2020 at 12:33:37PM -0300, ?talo Augusto Cavini wrote:
> Dear all,
> 
> I am in the deposition process and I am having considerable differences in
> R-factors calculated by DCC compared with Buster values. For one structure
> processed at 1.35 A, in Buster we have R=0.224 and Rfree=0.248 but in the
> validation report DCC gives R=0.233 Rfree=0.263. And the same does not
> happen for structures refined with other refinement tools, Phenix.refine
> for example.
> 
> By the way, I am using wwPDB OneDep System and PDB_Extract for validation
> and I am going straight from Buster output to wwPDB OneDep (no model
> changes).
> 
> Has anyone ever experienced something similar?
> Any insights about how solving this problem are appreciated.

to add to the thread that Ian referred to.

The R-values that BUSTER reports involve the expectation value of the
full "calculated" structure factor. This value is now written to the
final MTZ file in column FX. We also provide a tool, rvalue, that can
be used to compute R-values based on either FX or the more common FC.
For example you can try the following on the example from a refine
test[1]:

% rvalue -fx refine.mtz
...
     Resolution     ---------------- shell average --------------------
        [A]            Nwork  Nfree    Nall     Rwork    Rfree    Rall 
...
  24.848 -   2.621 :   19350   1669   21019    0.2104   0.2460   0.2132

or for FC (rvalue -fc refine.mtz):
...
  24.848 -   2.621 :   19350   1669   21019    0.2128   0.2493   0.2157


It is our understanding that DCC is using REFMAC to compute R-values
so differences in R-values can then also be attributed to different
bulk-solvent models etc.

Hope this helps.


Best regards,

ClAuS, Clemens and Gerard (for BUSTER Developers)
[1]: https://www.globalphasing.com/buster/wiki/index.cgi?BusterShortRefineTest2