2 Replies Latest reply on Oct 10, 2011 11:48 AM by 601256

    Read Across Saved Models

    601256 New User
    601256 Dec 14, 2011 10:23 AM

      Hi all,

      When carrying out a read across prediction the user is prompted to save the model. If this is done the model then becomes available to use as a QSAR model in the Gap Filling module. Does anyone know how this operates? In a limited exploration of this feature, all predictions I've managed to make using a saved read across model have been identical to the original prediction used to construct the model. I'm working my way through the prediction report (257 pages) and the model report (50 pages), but so far none the wiser. Does anyone know how this feature works and can be used for future predictions?

       

      Nick

      • 8703 Views
      • Categories: Data gap filling
      • Tags:
        • Read Across Saved Models
          83059 Apprentice
          83059 Oct 10, 2011 10:57 AM (in response to 601256)

          Hi Nick,

           

          I'm not sure Iunderstood your question.

           

          The function allows you to store the category so that you can later apply it to other chemicals, provided that they are in the applicability domain. It will always make identical predictions for your target chemical. For read-across categories it will indeed always come up with the same answer, but the answer will be true only if the chemical is in the applicability domain (i.e. the set of rules that you used to build your category).

           

          And that's where the crux lies.

           

          It makes little sense to save a category which is taylored precisely to your target chemical (i.e. has a very small applicability domain).

           

          But for categories which have a wider applicability domain, it might prove to be useful to save the model and reuse it for the prediction of other chemicals.

           

          Don't know if this is helpful. Maybe I misunderstood your question.

           

          cheers

           

          Bob