I'm looking at trying to build models of a set of membrane-bound
proteins via structural homology, and I was wondering if there are any
special caveats that apply to this situation.
Modeller will not treat such proteins specially, so whether it builds
reasonable models depends on whether they _need_ to be treated
specially. If you have reasonable sequence identity, it is likely to do
a good job at building models, since by construction your model will
look like the template where the two are aligned - but in low seq.id.
cases it has to rely more on its library of known protein structure
characteristics, and this is constructed using only globular proteins,
so may not accurately represent any special cases.