We're trying to model a membrane peptide of ~50 aminoacids which likely assumes beta-strand conformation in the membrane. Can we do this with the current Ab Initio membrane implementation of Rosetta? I know that we should use Octopus to generate a span file, but I'm left with the impression that it's designed for alpha helical proteins and that Rosetta in general is biased toward generation of alpha helices. Is there a way to enforce some beta-sheet restrictions during the ab initio folding process?
Most of the Rosetta membrane code assumes helical membrane proteins. (And particularly helicies which span the membrane completely.)
We're starting to move away from this formulation, with the recent MP-framework formulation (https://doi.org/10.1371/journal.pcbi.1004398). This can, theoretically, handle non-alpha helical membrane structures, though much of the setup code around span files still assumes such topologies. (See https://www.biorxiv.org/content/10.1101/630715v1.full for an example of the use of the MP framework with a beta-barrel protein.)
Even with the MP framework, you're likely to have issues with the sampling scheme for some protocols. The standard Rosetta MembraneAbInitio takes advantage of the fact that your typical membrane protein is a bundle of alpha helicies to do the sampling. It won't necessarily work all that well with a structure which doesn't have that sort of conformation. You may need to adjust the sampling approach you use based on what, exactly, you think the structure looks like in the membrane and how you want to sample things. (For example, if your protein is a beta strand in the membrane, what do you think is happening with the "edge" hydrogen bonders in the sheet? Leaving them unsatisfied is likely to be energetically unfavorable. Depending on how you want to pair them off may change how you do the sampling.)(