We plan to follow/advise the following steps for protein-protein docking when no known information on the complex structure is available and, it is necessary to do an initial global search around one partner.
The objective of clustering is to, ideally, seek larger clusters of decoy structures having more closely spaced energies.
Hence, the following sequence of steps is proposed for protein-protein docking where no known information on structure is available:
1. Generate many thousands of decoys using low_res_protocol_only in RosettaDock, and a global search over all space around fixed partner.
2. Cluster this set and select highest scored structure from 1st of 2nd highest ranked cluster (check structure visually).
3. Run high-resolution (docking_local_refine) RosettaDock with this selected structure as starting structure--generate 1000-10000 decoys.
4. Cluster the high-resolution set and, again, select highest scored structure from 1st tor 2nd highest ranked cluster.
Any comments/additions to this sequence, would be welcomed.