Predicting the part of protein strcutrue

You would fold them as separate domains – as this is how they typically would fold.  The Transmembrane domain would need to be folded using RosettaMembrane.  Search the docs and you should find quite a bit on it.  https://www.rosettacommons.org/docs/latest/application_documentation/structure_prediction/membrane-abinitio You could then fold the cytosolic part independantly.  I’m not sure if folding the cytosolic component with the transmembrane domain and an implicit membrane is feasible at this point.  You could surely try – as if the cytosolic domain has parts that are proximal to the membrane, it could help.  I would probably attempt that using RosettaScripts, but you would need to figure out what all to use and add the membrane info using the AddMembraneMover to your transmembrane domain.  The tricky part is adding a sequence on to an already partially folded protein.  For that, you could use RosettaCM to get the base pdb structure, and model the cytosolic domain. https://www.rosettacommons.org/docs/latest/application_documentation/structure_prediction/RosettaCM  I’m not sue if the AddMembraneMover and RosettaCM work well together, but they should.