[MMTK] protein sizes and other stuff

Konrad Hinsen hinsen@cnrs-orleans.fr
Wed, 10 Nov 1999 08:59:24 +0100


> Hi.  I've been attempting to read largish PDB files into
> MMTK with no success.  (I'm using 2.0b1 of both MMTK and
> ScientificPython.) The code fragment is almost exactly
> the one from the examples which preserves atom order:
> 
> config = PDBConfiguration('abcd.pdb')
> universe = InfiniteUniverse(Amber94ForceField()
> universe.addObject(configuration.createAll(None, 1)) 

That should work fine with all sufficiently standard-conforming PDB files.
Supposing of course that you replace "configuration" by "config" in the
last line.

> I've gotten different errors depending on whether or not
> I cut the PDB file to make it shorter.  We're talking
> about something that's 100+ amino acid residues.  The

No problem - I use files with thousands of amino acids!

>   File "/usr/lib/python1.5/site-packages/MMTK/PDB.py", line 298, in setResidueCo
> nfiguration
>     raise IOError, 'Atom '+atom.name+' of PDB residue ' + \
> IOError: Atom H of PDB residue LYS not found in residue Lys of object .Lys70

That error message indicates a mismatch between the atoms in the PDB
file and the atoms in the MMTK database definition for a residue.
Specifically, it claims that lysine should not have an atom called
"H", which is obviously wrong (the H is in the peptide group). Such
error messages should only occur for PDB files that use non-standard
atom names.

I don't know for sure what goes wrong here, but I suspect that your
MMTK installation is somehow broken, such that you are using wrong
database files. Could you send me a PDB file which produces such a
problem? I am rather sure that it will work for me, but I could then
help you to analyze the problem.

> Reading in something an order of magnitude smaller like a 7
> residue a.a. chain works fine.  Are there any statically dimensioned
> arrays that need to be resized?  (A la 'sizes.h' in the sander

No. Everything is allocated dynamically. There is no place in MMTK
where a static array dimension could have to be adapted to the system
size.

> Some other stuff -- does anyone know how fast MMTK is
> is compared to other programs such as sander (in the amber suite)?

That depends very much on your system, your computer, and your
compiler!

> I'm interested doing an md run of a 27,000 atom system (with most of 
> the atoms being part of water.)  Which leads me to another question:
> Does MMTK use the TIP3P water representation?  I haven't been
> able to find this anywhere.

As long as you use the Amber 94 force field, you get the Amber 94
water model, which is a modified TIP3 (the original TIP3 is rigid, the
one in Amber is flexible). But this depends entirely on the force
field; I have done simulations of SPC/E water just by using a
different force field.

Konrad.
-- 
-------------------------------------------------------------------------------
Konrad Hinsen                            | E-Mail: hinsen@cnrs-orleans.fr
Centre de Biophysique Moleculaire (CNRS) | Tel.: +33-2.38.25.55.69
Rue Charles Sadron                       | Fax:  +33-2.38.63.15.17
45071 Orleans Cedex 2                    | Deutsch/Esperanto/English/
France                                   | Nederlands/Francais
-------------------------------------------------------------------------------