Can I change high-resolution filter in SymDock

Member Site Forums Rosetta 3 Rosetta 3 – Applications Can I change high-resolution filter in SymDock

Viewing 1 reply thread
  • Author
    Posts
    • #2314
      Anonymous

        I am doing SymDock using Rosetta 3.5. I played with dock_lowres_filter so decoys were generated. Now the program reports that 

        protocols.symmetric_docking.SymDockProtocol: STRUCTURE FAILED HIGH-RES FILTER

        and no decoys could be generated. Unlike the failures with lowres_filter, the precise values are not printed for comparison for high-res filter. Is there a way for me to tune high-res filter like dock_lowres_filter ? Or is it generally a bad idea to touch it?

         

      • #11282
        Anonymous

          There’s two things going  into the  highres filter, the total (fullatom) score and the interface score (I_sc). The cutoff for the first is controlled by  the -cluster:output_score_filter option, and the second is hard coded at 0.

          Recent versions of the SymDockProtocol (the weekly releases since mid 2014) will actually print out the two values that it encountered on the error line. Rosetta3.5 doesn’t have that, though. 

          The defaults for both of these are rather generous, though. By default the fullatom filter is set to 1000000.0, which means anything but the most massively broken structure is likely to pass. (Unless you’re doing something strange with structure scoring.)  And if the interface score is greater than zero, that means that Rosetta doesn’t think things aren’t going to bind. If you’re not passing the default high res filter, things are probably going wrong with your docking job.

          There is a very limited facility to turn off the filtering, though. If the -docking::dock_ppk flag is set, the high-res filter will be skipped – though that will also change how docking progresses, so I wouldn’t suggest it as a general solution.

        • #11283
          Anonymous

            deleted

             

        Viewing 1 reply thread
        • You must be logged in to reply to this topic.