On 6/30/10 11:43 AM, Diane Lynch DLLYNCH wrote: > We have been using the loop modeling functionality (loopmodel class of > modeller, mostly 8v2 but recently upgraded to 9v8) to add loops onto our > receptor. ... > The script is below and it runs sucessfully. From this i can generate > loops that seem to recognize that the ligand is there

This is probably just luck - the loopmodel protocol does not consider loop-ligand interactions (other than soft-sphere clashes - see below), so the loop could certainly interact poorly with the ligand (e.g. it won't preserve protein-ligand covalent bonds or other strongly directional interactions). You could avoid this by overriding the special_restraints() method to add whatever loop-ligand restraints you think are appropriate.

> 1)How (using default settings, for example in the script below) does > modeler decide what van der waal radii to assign the atoms in the ligand?

BLK ligands have no topology and no parameters, so Modeller has to guess based on the name of each atom - for example it will guess that an atom labeled 'C1' is a carbon and use the carbon radius defined in modlib/radii.lib.

> I read about the loops and electrostatics/H's in > http://salilab.org/archives/modeller_usage/2005/msg00244.html which is > pretty old, so I am not 100% certain things have not changed

That information is still accurate for the loopmodel protocol.

> 2)Given that the loop modeling routine uses a statistical potential, and > is not necessarily parameterized to work well with > electrostatics, does it assign default charges to the ligand?

No - Modeller does not use any kind of electrostatic potential, so even if it did assign charges they wouldn't be used anyway.

> 3)Given that the loop modeling routine is not parameterized for H's. Is > it better to remove the H atoms from the ligand as well?

I don't think it'll make a lot of difference in practice, since the soft sphere radius of a hydrogen is pretty small, but you could certainly try that.

Ben Webb, Modeller Caretaker