[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [sdpd] GUI Interfaces - was Re: Ask for reports on ECM19
>A good Graphical User Interface can greatly assist in ensuring
>the starting file is what the user intended. (the most
>error prone option most likely being the insertion of
>fragments into the solving routines?).
If you mean that no error is possible with a GUI, I must
disagree. Verifying each parameter, one by one, twice,
with any text editor, will be as efficient as verifying each
parameter, one by one, twice, with a GUI. Where is the
difference ? People have to be careful with GUI programs
as well as with command-line programs. You may add a
GUI on DICVOL, but if you make a mistake on one peak
position, you will not index because the program has no
tolerance about unindexed peaks. No error test can tell
you that you entered a wrong data, GUI or not GUI.
Just a little story about an error still possible nowadays.
The SHELX76 and also SHELXS86, SHELXL/S93, 97 do
not generate automatically the symmetry coordinates, you have
to write them manually (why is a mystery for me, when so good
automatic space group interpreters exist for a while). A very common
error was to write 1/2 instead of 0.5, and the program did
not protest - you just could not solve the structure... If your
GUI ask for the symmetry coordinates and has not any test
about the fact that 1/2 should be given as 0.5, then you will
fail too.
About molecule location programs, being last-chance programs,
they may see particularly complex problems to solve, so that the
degree of success may be low. However, they may also see problems
unsolved simply because of errors in the previous steps (false cell,
false space group), and in these cases, they will fail too, GUI or not
GUI. The most error prone option should not be the insertion of
fragments in the model, because this is usually done by a copy paste
from some database (CSD or WebMolecules). And even the basic
GUI in ESPOIR shows the structure, allowing to detect starting model
errors without waiting for days. Moreover, molecule location is fast
if you use correctly the program : the 50 first reflections are enough
for locating one molecule, whatever big it can be, and the time needed
for 100000 rotations-translations will not exceed 15 minutes on a 1GHz
PC - with ESPOIR, one night for 50 independent Monte Carlo tests.
The program cannot find any good position for a wrong model in a
wrong cell or in the wrong space group.
Best
Armel Le Bail
http://www.cristal.org/course/
-------------------------- eGroups Sponsor -------------------------~-~>
Dial 1-800-555-TELL -- You Won't Believe Your Ears!
For more details, click here:
http://click.egroups.com/1/9537/14/_/40740/_/970558096/
---------------------------------------------------------------------_->