-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Hello David--
> I have been using the Python interface to calculate initial structures
> of my protein. I have used a slightly modified version of the
> anneal.py scripts given on the Xplor-NIH website and in the
> tutorials. However, I have made no significant changes to the loop
> that calculates one structure (structLoopAction or calcOneStructure).
>
> While the script appears to run fine, I've noticed that the '.sa' and
> '.viols' files are written before the final minimisation step in Cartesian
> space. As a result, there are discrepancies between the final energy and
> violations and those which are written to file. To troubleshoot, I've put in
> a line of code that writes a pdb file in the appropriate place and again, it
> writes the structure before completion of the energy minimisation.
>
I'll clarify the behavior of scripts using this template:
def calcOneStructure( structData ):
# [ do structure calculation ]
structData.writeStructure(potList)
from simulationTools import StructureLoop
StructureLoop(numStructures=100,
averagePotList=potList,
averageContext=FinalParams(rampedParams),
averageFitSel="name CA",
averageFilename="ave.pdb",
structLoopAction=calcOneStructure).run()
Structures are written out using structData.writeStructure()- no
optimization is performed on these structures after that line. However,
StructureLoop reads the structures back in and fits them (using
averageFitSel) and writes them back out. Because the PDB format only
contains three digits after the decimal point, this will lead to
slightly different energies if you re-evaluate the terms on the final
written-out files. Final statistics are determined by re-evaluating the
potential terms for each file- again there is roundoff error because of
this, so the energies/violations calculated are slightly
different. Finally, a minimization procedure is performed using
averagePotList with parameters from FinalParams(rampedParams).
If you find behavior different than this, there is a bug somewhere. If
you can upload a complete tarred example to
http://nmr.cit.nih.gov/xplor-nih/download.cgi
I will look into it.
best regards--
Charles
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.10 (GNU/Linux)
Comment: Processed by Mailcrypt 3.5.9 <http://mailcrypt.sourceforge.net/>
iEYEARECAAYFAkwaMsEACgkQPK2zrJwS/laFBACfWvdpsIaLCu26yEb72isD3y1s
MjIAnjO6iM3I7rm7nM3rEqwVLG6Y08TQ
=ZcPI
-----END PGP SIGNATURE-----
