CCL: Molecular descriptor selection

 Sent to CCL by: Isaac B Bersuker [bersuker(-)cm.utexas.edu]
 One possibility (rigorous, but not easy) is to employ QSAR without arbitrary
 descriptors, see J. Comp.-Aided Mol. Des. 22, N 6-7, 423-430 (2008).
 Dr. Isaac B. Bersuker
 Institute for Theoretical Chemistry
 The University of Texas at Austin
 Chem & Biochem Department
 1 University Station A5300
 Austin, TX 78712-0165
 Phone: (512) 471-4671; Fax: (512) 471-8696
 E-mail: bersuker- -cm.utexas.edu
 http://www.cm.utexas.edu/isaac_bersuker
 ----- Original Message -----
 > From: "Sangeetha Subramaniam srdshigella .. gmail.com"
 <owner-chemistry- -ccl.net>
 To: "Isaac B.  Bersuker" <bersuker- -mail.cm.utexas.edu>
 Sent: Wednesday, September 9, 2009 7:46:35 AM GMT -06:00 US/Canada Central
 Subject: CCL: Molecular descriptor selection
 Sent to CCL by: "Sangeetha  Subramaniam" [srdshigella],[gmail.com]
 Hello everyone,
 I have a query and would be glad to hear all your suggestions.
 While building a model for a set of compounds, how does one make the choice of
 molecular descriptors. As more than thousands of them can be calculated, it
 looks good to use all of them. But will it make sense?
 What methods/packages can be used in deciding the right choice of descriptors?
 Can you please post relevant software that can be used here..
 Thanks
 Sangeetha.http://www.ccl.net/cgi-bin/ccl/send_ccl_messagehttp-:-//www.ccl.net/chemistry/sub_unsub.shtmlhttp-:-//www.ccl.net/spammers.txt