Hi, Francois,

A MODELLER TOP script can relatively easily iterate over reading a pdb file, doing an energy calculation, and printing out the results, for many input pdb files. But the problem is getting the "true" CHARMM energy out of MODELLER.

1) The CHARMM terms in the MODELLER objective function are only a subset of all the CHARMM terms (there are only bond length, bond angle, and impropers in the default MODELLER objective function constructed for homology modeling, though one could add the dihedral angle, Lennard-Jones, and Coulomb terms by selecting the right options; I think CHARMM also uses some fancy cross-term bond angle terms which are not implemented in MODELLER).

2) The CHARMM terms are strictly speaking meant for the all-atoms models (including hydrogens), so I think one should build hydrogens before assessing MODELLER models with CHARMM.

3) MODELLER does not have the latest CHARMM force field.

4) You probably need to minimize a MODELLER model in CHARMM, just a little bit, before evaluating its energy (otherwise it would "artificially" high).

So, I think it would be best if you do the energy calculations in CHARMM. Also, there is a web server (EEF1 server or so ...) here in NYC by Themis Lazaridis, which may give you want you want (maybe not for thousands of structures?)?

Hope things are going very well for you!

Best wishes, Andrej

-- Andrej Sali, Associate Professor The Rockefeller University, 1230 York Avenue, New York, NY 10021-6399 tel +1 212 327 7550; lab +1 212 327 7206 ; fax +1 212 327 7540 e-mail sali@rockefeller.edu; http://salilab.org

-----Original Message----- From: owner-modeller_usage@salilab.org [mailto:owner-modeller_usage@salilab.org] On Behalf Of Francois Sampieri Sent: Wednesday, January 16, 2002 1:08 PM To: modeller_usage@salilab.org Subject: Potential energy calculations with MODELLER

Hi and Greetings to all !

I used MODELLER to generate a huge number (3500 pdb files) of models of the same protein with 7 different disulfide pairings (500 models of each pairing), and now I have to sort the models according their potential energies (bonded + non-bonded) in vacuum. I initially thought of using a molecular mechanics software in standalone mode to be lauched on each pdb file by a perl script, and then collect the energy values in whatever format the software created. I start looking for such a program which should be: - executable for SGI 02 R10K (IRIX 6.5); - free for academics; - launchable from a Perl script; But I didn't found such a jewel (oh no!, not such of highest value as MODELLER!) on the Web. So I come back to MODELLER, though its top script mode didn't seem the simplest way to me to achieve batch energy calculations, but I hope I'm wrong.

Could anybody suggest a method to use MODELLER for computing (CHARMM) potential energy of billions of structures in pdb format?

Many thanks in advance!

François Sampieri

____________________________________________ UMR 6560 CNRS-Universite de la Mediterranee Laboratoire de Biochimie - IFR Jean Roche Faculte de Medecine Nord Bd Pierre Dramard 13916 Marseille Cedex 20 E-mail: sampieri.f@jean-roche.univ-mrs.fr Fax: (33) 4 91 65 75 95 Tel: (33) 4 91 69 88 44