[phenixbb] Fwd: ML with Twinning?
kellerj at janelia.hhmi.org
Wed Dec 16 12:40:25 PST 2015
There are a number of twinning tests, and if they indicate a ~>5% twin fraction, the convention is to use twin refinement. Refining twin fractions then gives a more accurate estimate of the true fraction.
I think your hypothetical about not helping anything besides R might not really happen in practice—has it happened to you, or do you know of such a case? In any case, if you know something is there in your data (twinning, waters, chloride ions, etc) why not model it if you have reasonable evidence that it’s there?
From: phenixbb-bounces at phenix-online.org [mailto:phenixbb-bounces at phenix-online.org] On Behalf Of wtempel
Sent: Wednesday, December 16, 2015 3:21 PM
To: PHENIX user mailing list
Subject: [phenixbb] Fwd: ML with Twinning?
please allow me to use this prompt for my twinning-related question.
A recent version of xtriage prints this warning:
It might be worthwhile carrying out refinement with a twin specific target function.
Please note however that R-factors from twinned refinement cannot be directly
compared to R-factors without twinning, as they will always be lower when a
twin law is used. You should also use caution when interpreting the maps from
refinement, as they will have significantly more model bias.
Consider a case where specification of a twin law produces a “significant” reduction in the residuals, say between 5 and 10%-points. Maps have not revealed any additional features or model errors. Model geometry (such as fraction of residues in favored regions of the Ramachandran plot) has not improved. Should I specify the twin target during refinement?
How do my colleagues decide when to use twin refinement?
————— Forwarded message —————
From: Pavel Afonine pafonine at lbl.gov<http://mailto:firstname.lastname@example.org>
Date: Wed, Dec 16, 2015 at 2:06 PM
Subject: Re: [phenixbb] ML with Twinning?
To: “Keller, Jacob” kellerj at janelia.hhmi.org<http://mailto:email@example.com>, “phenixbb at phenix-online.org<mailto:phenixbb at phenix-online.org>” phenixbb at phenix-online.org<http://mailto:firstname.lastname@example.org>
Is Phenix able yet to use the ML target function with twinned data?
Is it in the works?
There are formulas written out:
“Maximum likelihood refinement for twinned structures”:
some one needs to code it.
phenixbb mailing list
phenixbb at phenix-online.org<mailto:phenixbb at phenix-online.org>
Unsubscribe: phenixbb-leave at phenix-online.org<mailto:phenixbb-leave at phenix-online.org>
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the phenixbb