[phenixbb] r-work and r-free for specific resolution range(s)

Kay Diederichs kay.diederichs at uni-konstanz.de
Wed Aug 4 12:51:12 PDT 2010


> Date: Wed, 4 Aug 2010 10:37:45 -0700
> From: Nathaniel Echols<nechols at lbl.gov>
> To: PHENIX user mailing list<phenixbb at phenix-online.org>
> Subject: Re: [phenixbb] r-work and r-free for specific resolution
> 	range(s)
> Message-ID:
> 	<AANLkTindW12Anq8iJVq-rh5LWiUyj4rZ52M3iyoy=7j8 at mail.gmail.com>
> Content-Type: text/plain; charset="iso-8859-1"
>
> On Wed, Aug 4, 2010 at 10:22 AM, Kay Diederichs<
> kay.diederichs at uni-konstanz.de>  wrote:
>
>> is there a way to get the R-factor statistics from phenix.refine, for a
>> specific resolution range, e.g. 2.5-2.6A ? The reason I ask is that I would
>> like to specify the same resolution range as the data reduction program
>> does.
>>
>
> phenix.refine will print R-factors by resolution shell, but the number of
> shells it decides to use varies with the number of reflections - for a
> really small structure (e.g. amyloid peptide) you will have only one
> resolution shell (the argument being that R-free statistics aren't reliable
> if you have very few reflections in the test set).  You can set an upper
> limit on this (refinement.main.max_number_of_resolution_bins), but it won't
> give you ten shells if phenix.refine wants to use, say, five.  There may be
> another way to get this information that I'm not aware of, but I've always
> done it directly in Python.
>
> I've been working on a project that potentially requires changing
> phenix.model_vs_data to report these statistics; my plan was to use the last
> of ten shells, since both HKL2000 and SCALA report everything that way
> (don't know about XDS since I haven't used it).  Is this sensible, or does
> it need to be more flexible and accept explicit resolution ranges?  (I'm not
> actually sure whether this is possible right now, but 10 shells is easy.)
>
> -Nat

Hi Nat,

in fact this is a small protein, and phenix.refine only outputs a single 
resolution shell.

phenix.model_vs_data currently does not give R-factors; it would be 
worthwhile to enhance this, and to have it accept shell limits.

XDS has 9 shells whose limits are calculated internally, and one 
overall. For XSCALE, one can specify up to 20 RESOLUTION_SHELLS .

In the meantime, I resort to

sftools <<EOF
read x_f_model.mtz
Y
select col R_FREE_FLAGS=0
correl col FOBS FMODEL resolution 2.6 2.5 shells 1
select all
select col R_FREE_FLAGS=1
correl col FOBS FMODEL resolution 2.6 2.5 shells 1
stop
y

where x_f_model was written by phenix.refine .

thanks,

Kay

-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 5236 bytes
Desc: S/MIME Cryptographic Signature
URL: <http://phenix-online.org/pipermail/phenixbb/attachments/20100804/3cee34e1/attachment-0003.p7s>


More information about the phenixbb mailing list