Mike White wrote: > I have modeled a homopentamer using automodel with enforcing pentameric > symmetry, and everything looks fine. There are two regions of each > subunit that have 3 and 5 residue long loops that aren't in the > templates. When I do loop modeling of all 10 of these loops at the same > time using a script based on the one in the Advanced tutorial, I get a > set of models where the loops are no longer symmetric. > I then introduced the same symmetry constraints in the loop modeling > that I used in the original model, and it produced models with the > appropriate symmetry. However, it took much longer (765 seconds of cpu > time to generate 5 loop models without the symmetry , but 8629 seconds > to generate 2 loop models with symmetry). I don't mind letting things > run overnight to generate models, but the hugh increase in computational > time suggests that I might not be doing this right.
Your script looks fine to me. Symmetry restraints are certainly going to be expensive to calculate, because they require calculation of all interatomic distances, d(ij). See http://salilab.org/modeller/manual/node251.html. The loop modeling potential, on the other hand, only works on nonbonded interactions, which are limited by the nonbonded pair list and exclude bonded atoms.
One way you can speed this up is to only constrain a subset of atoms. For example, you could constrain the backbone, or even just the C-alpha atoms, to be symmetrical. Simply change atom_types from 'ALL' to 'MNCH' or 'CA'.
Ben Webb, Modeller Caretaker