On Fri, 27 Aug 2004, Kevin Karplus wrote: > >Yes, the family pairwise search by Bill Grundy (now William Noble) is >an interesting and fairly simple technique for trying to improve >homology detection by using known relationships among sequences in the >database. I've had students implement Grundy's "product of p-values" >technique for our HMM-based scoring methods. In their preliminary >tests, they showed that it might help, but their projects were never >taken to the point where the method could be incorporated into an >automatic prediction method. > >One problem with that method is that it treats all templates in a fold >class as equally informative and does not use information from >templates in competing fold classes. One can envision combining >methods such as logisitic regression that could use the extra >information to get better predictions. One danger is that >multi-domain proteins correctly have multiple correct folds, so simple >competition between fold classes is not quite the right model. Yup, I was thinking of implementing something like this along the lines of the 'alignment universe' data structure implemented by Liisa Holm and Andreas Heger (sp?). This way you could combine domain boundary information into the calculation and additionally build up a domain classification hierarchy from the ground up (using the information provided by the network of homologus relationships). I never got very far along these lines either. Cheers, Dan. > >------------------------------ >Kevin Karplus karplus at soe.ucsc.edu http://www.soe.ucsc.edu/~karplus >Senior member, IEEE Board of Directors, ISCB (starting Jan 2005) >Professor of Biomolecular Engineering, University of California, Santa Cruz >Undergraduate and Graduate Director, Bioinformatics >Affiliations for identification only. > > >_______________________________________________ >ssml-general mailing list >ssml-general at bioinformatics.org >https://bioinformatics.org/mailman/listinfo/ssml-general >