I did perturbative docking using both rosetta2.3 and rosetta3.1, which gave me very different score vs. rmsd results. The rosetta2.3 results showed slight funnel in score vs. rmsd relationship, while the rosetta3.1 gave me results with no such kind of tendency at all. Anyone having similar experience with the new edition?
Following is the flags and commands comparison in rosetta2.3 and rosetta3.1:
rosetta2.3: -dock -pose -dock_mcm -dock_rtmin -quiet -nstruct 5000 -fake_native -randomize1 -randomize2 -ex1 -ex2aro_only -unboundrot
rosetta3.1: -packing:ex1 -nstruct 5000 -docking:spin -docking:randomize1 -docking:randomize2 -docking_local_refine
I tried to make the rosetta3.1 flags consistent with rosetta2.3, but some of the flags could not be turned on. any suggestions on how to improve the rosetta3.1 results? Thanks a lot.
The flags are set differently in the two command lines, the 3.1 command line that best replicates the 2.3 command line you provided is is:
-ex1 -ex2aro -nstruct 5000 -fake_native -randomize1 -randomize2 -unboundrot
dock_rtmin is not currently supported in 3.1; docking_local_refine will lead to significantly different sampling as the entire low-res stage will be skipped.
Also check to see if the 'slight' funnel in 2.3 is reproducible in 2.3 and not the result of stochastic noise.
We are currently in the process of benchmarking the 3.1 docking performance with the 2.3 docking performance, and will make the results public when we have them.