[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: DURBAN-2003

• To: Multiple recipients of list <ecacomsig@iucr.org>
• Subject: RE: DURBAN-2003
• From: "David Watkin" <david.watkin@chemistry.oxford.ac.uk>
• Date: Fri, 13 Sep 2002 10:06:50 +0100 (BST)

hello
	Here are my proposals - sorry they are late but I have been at a
workshop etc since returning from Geneva.

1 I support the proposal for a session on Automatic Structure Solution.
See, for
example Rollett, Crystallographic Computing ed Ahmed, 1969, 'Methods have
been
developed for automatic solution of crystal structures'.  Dream on!   From
the small
molecule community possible speakers are Spek or Cascarano.  I don't think
the
powder people are any where near this, unless you include DASH (bill david)
which
attempts to fit a known molecular structure to the powder pattern.
macromolecular
people must have to tricks to share with the small molecule people.

2. Diffraction image processing.  At Glasgow I accidentally found myself in
a protein
session, where Major was giving an excellent presentation on how Denzo
works.
Lots of small molecule people missed it.  Duisenberg (?) gives an excellent
talk on
shoebox fitting (EVALCCD).  Something to reassure the users about the
virtues ( and methods) in using spherical harmonics to model anisotropy
(absorption) in the diffraction data (SADABS, SORTAV, SCALEPACK)

3 Something about data collection itself.  The Gorbitz paper (Acta Cryst,
about 1999)
shows that massive crystals yield acceptable data.  We have convinced
ourselves that
collecting data for a long time does not necessarily yield better data.  We
also have
looked at improvement vs temperature for a few compounds. Im sure some one
else
has done the same and could speak about it.  Some of the results are not
intuitive.

4. Algorithms.  Dale Tonrud does an excellent talk showing how the several
refinement algorithms are related (from CG to ML).  Does anyone know
anything
about Least Moduli as a method?  I read a paper about 10 years ago (not
crystallographic) which claimed it was better than sliced bread.  Something
on how to
write your own Genetic Algorithm  (Simulated Annealing, Monte Carlo etc)
program.
These methods are still fashionable and even work. ( see Tam, K. Y. &
Compton, R.
G. (1995). GAMATCH - a genetic algorithm-based program for indexing crystal
faces. J. Appl. Cryst. 28, 640-645).  Some one to give an example of one
successful
piece of research completed thanks to Neural Networks.  I know of one
successful
application (at least, that's what the author claims) but its not
crystallography.

5. Validation (as ever!).  Spotting that something might be wrong has become
a fine
art, but it still needs a crystallographer to check if something is in fact
wrong, and to
suggest how it might be put right.  How wrong does wrong have to be before a
paper
is rejected?  Links back to item 3.  Distinguish between wrong and poor.  Is
a poor
structure better than no structure?

Best wishes
				David

-----Original Message-----
From: ecacomsig@iucr.org [mailto:ecacomsig@iucr.org]On Behalf Of
a.l.spek
Sent: 11 September 2002 10:47 am
To: Multiple recipients of list
Subject: DURBAN-2003


Dear Computing SIG Members,

	I received the following suggestion for a computing/teaching
session:

"Modern programming languages and programming techniques: can they
 lead to higher reliability and programmer productivity"

Best wishes

Ton Spek

Chairman of the Computing SIG

Reply to: [list | sender only]

• Prev by Date: Re: Sept 15 Deadline for proposals.]
• Next by Date: RE: DURBAN-2003
• Prev by thread: Re: DURBAN-2003
• Next by thread: RE: DURBAN-2003
• Index(es):
  • Date
  • Thread