Dear Jeffrey,
just using thin shells of reflections for the test set might not be
enough to eliminate any correlation to working set reflections. The
distance of the test set reflections in reciprocal space, after applying
the NCS rotations in reciprocal space, to any reflections of the working
set should be large enough to be uncorrelated. A typical maximum
distance for correlation in reciprocal space is the first zero (root) of
the Fourier transform of the approximated spherical shape of the protein
molecule, the so-called G-function (see [1) and [2]). However, this
means, that shells should be thicker than 2 times this distance to
blindly exclude any potential correlation between test set and working
set reflection in reciprocal space. This would be way too expensive!
An better way to select uncorrelated reflections is described in [3]
(same reference that Kay has given), where for each test set reflection,
after applying the NCS rotations in reciprocal space, the distances to
the working set reflections are taken into account, and test and working
set reflections are chosen such that their distances are large enough to
be uncorrelated.
Best regards,
Dirk.
[1] Rossmann & Blow, Acta Cryst, 15, 24 -31 (1962)
[2] Main & Rossmann, Acta Cryst, 21, 67-72 (1966)
[3] Fabiola, Korostelev & Chapman, Acta Cryst, D62, 227-238 (2006)
On 8/27/21 4:47 PM, Kay Diederichs wrote:
Hi Jeffrey,
good question. Both twinning and NCS may couple reflections across free and
working sets, and this should be avoided by proper selection - otherwise Rfree
is biased towards Rwork. Selecting thin shells should be a good option, and can
be done in SFTOOLS (or DATAMAN, or SHELXPRO).
How much does it matter? Actually I started to search the literature after
reading your question, and expected that one of Z. Dauter's papers would
enlighten me, but until a minute ago couldn't find the one(s) that I thought
existed. But I found Fabiola et al., Acta Cryst. (2006). D62, 227–238 which is
relevant for the coupling by NCS. Ah I just found Smietanska et al., Acta
Cryst. (2020). D76, 653-667 where they indeed selected thin shells.
Best wishes,
Kay
On Fri, 27 Aug 2021 13:24:05 +0000, Jeffrey B Bonanno
<jeffrey.bona...@einsteinmed.org> wrote:
Hi Kay,
Can you also comment on Rfree set selection? It seems thin shell might be
preferred in these cases?
jbb
Jeffrey B. Bonanno, Ph.D.
Department of Biochemistry
Albert Einstein College of Medicine
1300 Morris Park Avenue
Bronx, NY 10461
off. 718-430-2452 fax. 718-430-8565
email jeffrey.bona...@einsteinmed.org
########################################################################
To unsubscribe from the CCP4BB list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1
This message was issued to members of www.jiscmail.ac.uk/CCP4BB, a mailing list
hosted by www.jiscmail.ac.uk, terms & conditions are available at
https://www.jiscmail.ac.uk/policyandsecurity/
--
*******************************************************
Dirk Kostrewa
Gene Center Munich
Department of Biochemistry, AG Hopfner
Ludwig-Maximilians-Universität München
Feodor-Lynen-Str. 25
D-81377 Munich
Germany
Phone: +49-89-2180-76845
Fax: +49-89-2180-76998
E-mail: dirk.kostr...@lmu.de
kostr...@genzentrum.lmu.de
WWW: www.genzentrum.lmu.de
*******************************************************
########################################################################
To unsubscribe from the CCP4BB list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1
This message was issued to members of www.jiscmail.ac.uk/CCP4BB, a mailing list
hosted by www.jiscmail.ac.uk, terms & conditions are available at
https://www.jiscmail.ac.uk/policyandsecurity/