Viewing 1 reply thread
  • Author
    • #2065

        Hi all,

        I’m trying to change the filters for the docking protocol.
        I tried :
        dock_lowres_filter 15 5
        This option is supposes to set manually the interchain_contact and the interchain_vdw cutoff.

        However my log file still says that the cutoff have the default value 10 1.
        protocols.docking.DockingProtocol: Scoretype: interchain_contact score: -12.5, cutoff: 9.99
        protocols.docking.DockingProtocol: Scoretype: interchain_vdw score: 1.30527, cutoff: 1

        I obtained this by running the docking_full_protocol (in rosetta_tests/integration/tests). I kept everything but the flags. I commented no_filters and add dock_lowres_filter 15 5.

        Thank you for your help.


      • #10577

          It does look like you are correct that those cutoff values are hard coded for the regular docking protocol. (They do look to be observed in the symmetric docking protocol.) From what I can tell, they haven’t been functional for the regular docking protocol since at least Rosetta3.3.

          Unfortunately, there doesn’t seem to be a way of changing the values short of editing the code and recompiling.

        • #10578

            Thank you very much for your answer.

            I may try to edit and recompile.
            Do you know what do those cutoff do?
            Looking at the log file, I saw that structures under the cutoff are rejected. But what does having a interchain_contact (interchain_vdw) of 10 (1) mean ?

            Thanks a lot


          • #10579

              The lines you’re interested in changing are in src/protocols/docking/ at around line 60. If you want, you could make it responsive to the commandline parameter using src/protocols/symmetric_docking/ around line 510 as a guide.

              The interchain_contact and interchain_vdw are lower-better (energy-like) metrics, so the filters should actually be rejecting poses with score higher (worse) than the cutoff. That is, at a interchain_vdw cutoff of 1.0, you should be keeping structures with a score of 0.5, but would reject structures with scores of 2.0 or 5.4.

              As to the meaning of the values, interchain_contact is a metric based on the number of residues in the interface. The more residues in the interface, the more negative the score is. The calculation is a little bit opaque, but it looks like for a two-protein complex in the standard docking protocol, a score of 10 would indicate that there is one or fewer residues in the interface. A score of 0 indicates 20 residues in the interface, and a score of -10 indicates that there is 40 or more residues in the interface. (This is from a quick read of the code – I have not confirmed this behavior, and the exact values will depend on how you run the protocol, etc. There may be a factor of two missing somewhere in there.)

              For interchain_vdw, it’s a metric based on the number of atoms which are clashing across the interface. Again, calculation is a little opaque, but it looks like the score should be zero if no atoms overlap, and you should get 1 for each atom pair across the interface which is right on top of each other. In practice, you’d never get exactly 1 for an atom pair, but the amount is greater the more overlap there is, and it’s cumulative, so you can get numbers greater than 1 for a large number of small overlaps, or a few atoms which overlap significantly. (Note that it is not a LJ-like calculation, and there’s no attractive component – it’s strictly repulsive, so favorable contact will not mitigate clashes elsewhere.)

            • #10585

                Thanks a lot for your helpful explanations.

              • #10612


                  I’m still working on these scores.

                  I used the docking_full_protocol on the tests repository.
                  I tried asking for multiple output with the flag -out:nstruct 1000, and I uncommented the no_filters flag.

                  In the score file, a few output (like #31, #33 and #35 ) have an interchain_contact score greater than 10.
                  I looked the source code of the protocol. This score does not seem to be evaluated after the filter.
                  How is it possible that it is greater than the cutoff?

                  Thank you for your help


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