[phenixbb] phenix.refine and R free values

Pavel Afonine pafonine at lbl.gov
Wed Sep 3 08:41:29 PDT 2014


Hi,

F(+),SIGF(+),F(-),SIGF(-) are derived from I(+),SIGI(+),I(-),SIGI(-) 
using various approximations that depend on underlying model. In this 
sense I(+),SIGI(+),I(-),SIGI(-) is more original and richer data than 
F(+),SIGF(+),F(-),SIGF(-).

If F(+),SIGF(+),F(-),SIGF(-) were used in refinement to obtain deposited 
structure then this would be a good reason to have both, F and I.

Pavel

On 9/3/14 8:33 AM, PC wrote:
> Hi Pavel and Phenix users.
>
> Thank you for your reply.
>
> I see in 2y78.mtz (example I used below) has both in the MTZ:
>
>   F(+),SIGF(+),F(-),SIGF(-)
>   I(+),SIGI(+),I(-),SIGI(-)
>
> I am confused, why? One is sufficient isn't it?
>
> Thank you,
>
>
>
>     -----Original Message-----
>     *From:* pafonine at lbl.gov
>     *Sent:* Sat, 30 Aug 2014 09:45:58 -0700
>     *To:* patrick.cossins at inbox.com, phenixbb at phenix-online.org
>     *Subject:* Re: [phenixbb] phenix.refine and R free values
>
>     Hello,
>
>>     Hi Phenix users,
>>
>>     I was trying to experiment with phenix.refine using default
>>     parameters (2Y78.pdb) and I got the following message.
>>
>>     Multiple equally suitable arrays of observed X-ray data found.
>>
>>     Possible choices:
>>       2Y78.mtz:IOBS,SIGIOBS
>>       2Y78.mtz:I(+),SIGI(+),I(-),SIGI(-)
>>
>>     Please use refinement.input.xray_data.labels
>>     to specify an unambiguous substring of the target label.
>>
>>     Anyway, two questions related to this:
>>
>>     1) How does one pick of the two options above?
>
>     If your question is about the syntax:
>
>     phenix.refine model.pdb data.mtz xray_data.labels="IOBS,SIGIOBS"
>     or
>     phenix.refine model.pdb data.mtz xray_data.labels="I(+),SIGI(+"
>     (sub-string should work).
>
>     If your question is about choice anomalous vs non-anomalous data
>     to be used: I would use anomalous because it is a richer source of
>     information, you will get anomalous difference map, you can refine
>     anomalous f'&f'', etc.
>
>>     2) If R work/R free went from 0.1044/0.1142  to 0.1096/0.1200 is
>>     that bad or is it acceptable?
>
>     R-factors look similar to me. Systematic slight increase may mean
>     you need to adjust target weights or choose refinement strategy
>     more tailored to the quality of input data and model.
>
>     Pavel
>
> ------------------------------------------------------------------------
> Free Online Photosharing - Share your photos online with your friends 
> and family!
> Visit http://www.inbox.com/photosharing to find out more!

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://phenix-online.org/pipermail/phenixbb/attachments/20140903/b04ccde6/attachment.htm>


More information about the phenixbb mailing list