On Tue, Nov 16, 2004 at 02:08:00AM -0500, Youbin Tu wrote: > I am using MODELLER7V7 to build a ligand-enzyme complex, I created > the ligand and modified the library files. Everything worked well > except the ligand in the final result seems too weird, for example the > benzene is not planar and there are many bonds between 2 atoms which > should only have one bond theoritically.

For non-standard residues (such as ligands) Modeller can only do as well as the parameters you give it, just as with any molecular mechanics package. If the ligand looks weird, you've almost certainly got an incomplete parameter set (and Modeller may warn you about this, depending on the circumstances).

If you give Modeller no parameters, it will guess them from the current template structure. See http://salilab.org/modeller/7v7/manual/node30.html. In the case of benzene, this will restrain your C-C bonds to their distances in the original structure, but it probably won't enforce sufficient constraints to keep planarity.

If flexibility of the ligand isn't too important, you should use a 'block' residue and keep the original geometry unchanged. This is also covered at the above URL.

> I thought maybe it is my problem when building the ligand. But > when I tried to use GDP which already existed in the library file. The > final results still have the same problem.

You may have conflicting restraints in this case.

Alicia's reply is largely correct. Modeller writes only xyz information into output PDBs, and no connection data; this is why viewing packages have to guess at the connectivity and sometimes get it wrong. Modeller uses the CHARMM (not CHARMm) 22 parameter set, and the coverage of residues other than aminoacids is rather sparse.

Ben Webb, Modeller Caretaker