Dear Rosetta help,
The ectodomain of my membrane protein has 100 to 250 residues with known structure (for which I plan to use a rigid broker), then about 100 residues with unknown structure, followed by a short transmembrane domain. If I apply the Rosetta membrane ab initio protocol to this protein, will this be a good way to solve the ectodomain structure along with that of the TMD? Or would it be best to first use a loops protocol to model the ectodomain (as in the homology_modeling_with_end_extension example folder) before running the whole thing with the membrane ab initio protocol?
Thanks in advance,
The membrane abinitio and the soluble protein abinito programs use rather different protocols for assembling proteins. I think your approach should be guided by how tightly coupled you believe the soluble domain and the transmembrane domain are. If you believe the folding of the two is tightly coupled, then you probably should try using the membrane ab initio program. If they're separate (e.g. they're two independent domains that are simply connected by a tether) then you may be better off modeling them as two separate domains which are then brought together.
Hi Rocco, Thank you for your reply. What if part of the ectodomain is not tightly coupled to the TM domain but it locates to a shallow position in the membrane, i.e. it interacts with the interface between the water and the membrane. Would this kind of an interaction of the ectodomain with the membrane be represented in the membrane ab initio protocol? Or is the membrane only represented in as much as it interacts with domains which are strictly located within the membrane? Regards, Holly
Non-transmembrane interactions with the membrane aren't captured very well with the current versions of the Rosetta membrane scorefunctions. There should be some disfavoring of placing polar sidechains "within" the membrane, even for non-transmembrane regions, but handling these sorts of non-embedded interactions is less well tested than standard transmembrane helixes.
I'd say that I wouldn't use the presence of non-transmembrane interactions in how you decide to structure the protocol. I'd model things however they're easiest to model, and then incorporate the membrane interaction only at the end, when you refine the complex as a whole in the presence of the membrane scorefunction.
Only if you're having problems getting good models would I look into modeling these sorts of non-transmembrane interactions explicitly.