On 11/16/11 5:21 AM, Vasillis Myrianthopoulos wrote: > I am trying to build a model of a transmembrane receptor and I get serious > problems and obvious steric errors on the loops which connect the TM > helices. The alignment is correct on a biological and phylogenetic basis > however there are large insertions on the loops.

You will need to do loop refinement on these loops to have any hope of generating a sensible conformation. But it depends on what you mean by "large". Loop modeling is limited to 12 residue insertions or so - Modeller does not do protein folding.

> First, the loops are modelled as > packed between the TM helices and not as free moving segments that can > float in the solvent environment

Modeller has no consideration of solvation - that comes implicitly from the template. And the loop modeling potential is derived from known structures of globular proteins, not membrane proteins. So if you have an insertion (and thus no template) it's not going to do the right thing here unless you add additional restraints.

> in some cases the loops form > kind of knots which are obviously wrong.

Look at the .ini file - if your insertion is long the initial model probably looks pretty bad, with the protein chain passing through itself. Modeller may not be able to recover from this and may generate knotted structures, particularly if you don't build many models. You can give Modeller a better initial guess at the structure to ameliorate this - it doesn't need to be stereochemically correct, but you can at least massage it so that the backbone is roughly where you think it should be.

Ben Webb, Modeller Caretaker