Re: [modeller_usage] Block Ligand stereo chemistry problem

[Modeller Caretaker <>]

Andre Fischer wrote:
> I am modelling a Protein with a ligand. The Ligand comes from the pdb
> template as a blk. Modeller works fine and the protein model seems
> reasonable, except that the modelled ligand is mirrored along its length
> axis compared to the template ligand. Also BLK Residues are defined
> rigid. some stereo operations seem to be allowed. How can i get rid of
> those? I want the ligand only to act as a steric constraint.

The rigidity is maintained by imposing restraints on all atom-atom 
distances within the BLK ligand. (See the nonstd_restraints function in 
the automodel class, at modlib/modeller/automodel/automodel.py, for the 
Python code that generates these restraints.) Therefore, it is possible 
that if your starting structure is sufficiently distorted, these 
restraints may result in an inverted structure rather than the original 
(or in rare cases they may not be satisfied at all).

When Modeller builds a model, the starting structure is randomized a 
little to ensure that each final structure is different. By default, the 
BLK residues are also randomized. You can thus ensure that no inversion 
or mirroring occurs by changing this randomization procedure to leave 
BLK residues untouched. See the documentation of automodel.rand_method. 
By default this is set to randomize.xyz, which from 
modlib/modeller/automodel/refine.py you can see does:
def xyz(atmsel):
    """Randomize all coordinates"""
    mdl = atmsel.get_model()
    atmsel.randomize_xyz(deviation=mdl.deviation)

- i.e., randomizes all coordinates by automodel.deviation. One way to 
prevent BLK residues being randomized is to do something like

def my_randomize(atmsel):
    mdl = atmsel.get_model()
    atmsel = atmsel - atmsel.only_no_topology()
    if len(atmsel) > 0:
        atmsel.randomize_xyz(deviation=mdl.deviation)

m = automodel(env, ...)
m.rand_method = my_randomize
m.make()

This modified randomization procedure is actually taken from the 
development version of Modeller, and will be the default in the next 
release.

	Ben Webb, Modeller Caretaker
--
             http://www.salilab.org/modeller/
Modeller mail list: http://salilab.org/mailman/listinfo/modeller_usage