How to dock 3 missing domains onto a multi-domain protein

[Anonymous]

Dear Rosetta users,

I have a hetero-dimer that consists of a catalytic (1068 residues) and a regulatory subunit (724 residues) which is wrapped around it. The structure of the catalytic subunit has been solved, but only part of the regulatory exists. More specifically, the regulatory subunit consists of 5 domains (A,B,C,D,E), which are linked by loops of <=30 residues. Two adjacent domains, C and D, exist in the crystal structure in complex with the catalytic subunit, whereas the rest (A, B, E) have been solved sepparately (as intact proteins). Therefore I need to assemble the full-length regulatory subunit by comparative modeling, and predict the orientation of domains A, B, and E relative to the catalytic and regulatory subunit domains C and D (these 3 would be moved as a rigid body or sort of). Is it possible to do this with Rosetta? Any advice would be greatly appreciated. thanks in advance.
Thomas

[Anonymous]

This is a very complex problem. There is no single command line to do something like this. Experimental constraints are going to be useful, do you have any?

I would start with docking B and E individually to the main+CD complex. If you get a good funnel on either of those (or if you have experimental constraints), try loop modeling to see if the B or E placement allows for a reasonable loop. Iterate through this, adding one more domain at a time.

You could hack the FloppyTail code a bit to get it to treat the B-C or D-E loop as a tail, then place the B or E domain as a big rigid chunk on the end of the tail, to try to sample what orientations B or E might land in (then iterate out to A). I think docking is superior here.

[Anonymous]

Rosetta 3.2 includes some crystallographic tools, but I don’t know if they are compatible with the data you have. Look for the cryoEM or molecular replacement tutorials in the manual/download.

If those won’t help, but you have a vague idea of where domain E goes from the weak density, then you can bias docking by playing E next to CD/cat in the correct location, but only use the -spin flag on domain E in docking, and do perturbative rather than global docking. This should try all orientations of E, but leave it in approximately the same place on the CD/cat surface.

You could also write constraints to get the same effect.

[Anonymous]

With respect to experimental constraints, the authors state in the paper that domain E was included in the crystal but wasn’t traced in the map. I’m not sure if I can create an starting structure of regC-D-E/cat with comparative modeling, and fit the domain E structure into the sparse map during refinement. Does this make sense?

[Author]

Posts