I would like to dock a small protein (59 amino acids plus one 140 atom ligand) to a large trimeric complex (358 amino acids *per monomer*, plus eight 140-atom ligands). The total number of atoms in the system is nearly 21,000.
When I run a small test of the system using the following flags (we use Torque for queuing jobs) and generate 10 decoys per instantiation, each decoy takes between 1800 and 2400 seconds to complete (30-40 minutes).
I have experimental evidence which suggests a particular region of the trimeric complex (let's call it the "top"...) participates in binding this small protein (though the site itself is not precisely known) and note that the trimeric complex itself is relatively stable.
Is there some way to fix atoms or *not* scan rotamers for the ligands on the "inside" and amino acids making up the "underside" of the complex to somehow speed up the process?
Ideally, I'd only like to sample conformations of the small protein, it's ligand, and the top surface of the trimeric complex.
Would the liberal use of site constraints (as mentioned near the end of: http://www.rosettacommons.org/content/docking-protein-symmetrical-complex) help at all?
Note that I have already successfully pre-packed the complex (that took only 300 seconds...) and am starting from that structure for the test case above.
Any thoughts and advice are welcome!