Re: Thin shells

[Bart Hazes <bart@sherlock.mmid.ualberta.ca>]

*** unsubscribe by sending the message "unsubscribe ccp4bb" to   ***
***       ccp4bb-requests@dl.ac.uk  or  Majordomo@dl.ac.uk       ***

SFTOOLS can do this with the

RFREE SHELL

option. The major problem is that in many cases you will get very thin
shells if you want to have a decent number of them. The smaller the unit
cell the more serious the problem. Programs that use RFREE reflections in
resolution bins can also run into problems as there may be bins with only
WORK reflections. If you have proper pointgroup NCS then it may be better
to use the NCS relations to group NCS correlated reflections into the same
RFREE set. This is not without practical problems either but seems better
to me. An RFREE NCS option has been coded into the next release of SFTOOLS
but needs more testing.

The thin shell option is in use in two projects in the lab (one with
20-fold none proper pointgroup NCS) and seems to be better than just
random selection. So if you have serious NCS then give it a try. You can
pick up the program by anonymous ftp at:

mycroft.mmid.ualberta.ca

The program is in the pub/sftools directory

Pick up the sftools100997.tar.Z file and the more recent sftools141097.f.Z
source code. Alternatively, you can get executables for IRIS 5.3 and Linux
(sftools141097.sgi_exe.Z and sftools141097.linux_exe.Z, respectively)

Bart