On Sep 14, 2007, at 5:47 PM, Geoffrey Hutchison wrote: > > I think Noel's discovery of RD-Kit and their 3D coordinate > generation may make this point moot.
I'm not familiar with the algorithm used by smi23d, but from the description that it used SPE, I'm going to assume that it was building a distance bounds matrix using a set of rules and then embedding that matrix into 3D. The key to this type of approach is, I think, the quality of the bounds matrix that one generates first.[1,2] The speed of the embedding method (whether it's diagaonalization or SPE) is secondary, particularly for drug-like molecules. If you think SPE is something helpful and are bothered by the patent, read the SPE papers carefully, particularly the JMGM 2003 paper, and then look at this paper (which Dimitris actually cites): http://ieeexplore.ieee.org/xpl/freeabs_all.jsp?arnumber=554199 Note that this predates the SPE papers and the patent filing. I wouldn't use CCA for embedding molecules, but it could be useful for doing MDS on larger data sets. Hope this is helpful, -greg [1] Our rules aren't the best, so this could definitely be an easy area of collaboration for us. [2] The speed of the force field you use to cleanup the geometries, and you usually do need to clean them up, is also important. ------------------------------------------------------------------------- This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ _______________________________________________ Blueobelisk-discuss mailing list Blueobelisk-discuss@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/blueobelisk-discuss