Loading ...
Sorry, an error occurred while loading the content.
 

Re: [cnsbb] turning off Rfree

Expand Messages
  • Edward A. Berry
    I have done this. I think it involves changing the target for optimization to something that doesn t involve maximum liklihood. I don t remember if it was
    Message 1 of 3 , Sep 24 12:18 PM
      I have done this. I think it involves changing the "target" for optimization
      to something that doesn't involve maximum liklihood. I don't remember if it was
      also necessary to set the flags to zero or get rid of them from the nolonger.cv
      file, but I don't think so.

      Crystal VanderZanden wrote:
      > Hello,
      >
      > I'm trying to run minimize in cns, but I want to set it so it doesn't do an Rfree
      > calculation, and so it will use all of my data in the minimization process. I have a
      > small low-resolution dataset that doesn't have very many reflections, so I'm having to
      > sequester about 20% of my data in order to run the Rfree calculation. Does anyone know
      > how to turn the Rfree calculation off?
      >
      > Things I've tried:
      > -making the minimize.inp target_bins=1 and flagging a very small percentage of my data as
      > a test set
      > -telling minimize.inp that my test_flag = 2 so no reflections would be marked as test
      > -telling minimize.inp that my test_set=(none)
      >
      >
      > Thanks in advance for any advice!
      > Crystal
      >
      >
      >
    • Ed Pozharski
      Crystal, I assume that none of the things you tried have worked and you probably got some kind of error message or otherwise results were atrocious. You must
      Message 2 of 3 , Sep 26 8:52 AM
        Crystal,

        I assume that none of the things you tried have "worked" and you
        probably got some kind of error message or otherwise results were
        atrocious.

        You must have the test set if you are using ML target in CNS. The
        reason is that test set reflections are used to estimate sigmaA, and
        without them the refinement target cannot be constructed.

        Ed Berry is absolutely right that you need to use a different target.
        Keep in mind that every structure in the PDB refined with CNS uses mlf
        target. Other targets are implemented but never used.

        Whether it's OK or not to run refinement at low resolution without
        cross-validation may be debatable (e.g. if the model is first refined
        with it, is it OK to turn cross-validation off as a last step when the
        danger of overfitting is minimized), but what you want to do is unlikely
        to gather majority support. I expect most people would agree that doing
        refinement at low resolution without cross-validation is a dangerous
        proposition.

        If you are willing to try, then Refmac and more recently phenix.refine
        both allow refinement without cross-validation. Beware that if this is
        something you plan to publish you are very likely to run into serious
        opposition from reviewers.

        Cheers,

        Ed.

        On Tue, 2013-09-24 at 11:51 -0600, Crystal VanderZanden wrote:
        >
        > Hello,
        >
        >
        > I'm trying to run minimize in cns, but I want to set it so it doesn't
        > do an Rfree calculation, and so it will use all of my data in the
        > minimization process. I have a small low-resolution dataset that
        > doesn't have very many reflections, so I'm having to sequester about
        > 20% of my data in order to run the Rfree calculation. Does anyone
        > know how to turn the Rfree calculation off?
        >
        >
        > Things I've tried:
        > -making the minimize.inp target_bins=1 and flagging a very small
        > percentage of my data as a test set
        > -telling minimize.inp that my test_flag = 2 so no reflections would be
        > marked as test
        > -telling minimize.inp that my test_set=(none)
        >
        >
        >
        >
        > Thanks in advance for any advice!
        >
        > Crystal
        >
        >
        >
        >
        >

        --
        Edwin Pozharski, PhD, Assistant Professor
        University of Maryland, Baltimore
        ----------------------------------------------
        When the Way is forgotten duty and justice appear;
        Then knowledge and wisdom are born along with hypocrisy.
        When harmonious relationships dissolve then respect and devotion arise;
        When a nation falls to chaos then loyalty and patriotism are born.
        ------------------------------ / Lao Tse /
      Your message has been successfully submitted and would be delivered to recipients shortly.