Applicability range of QSAR models

["Igor Baskin" <baskin "at@at" qsar.chem.msu.su>]

         Dear CCL'ers,
         Now we (three other authors and me) are finishing preparation
 of an article dealing with rigorous mathematical study of the problem
 of defining applicability range of QSAR/QSPR models. Most frequently
 this problem arises when (i) developing QSAR models for noncongeneric
 data sets, (ii) designing bioactive molecules based on QSAR models,
 (iii) comparing advantages and disadvantages of "knowledge-based"
 QSAR/SAR approaches (aimed at making predictions through understanding
 molecular mechanisms) vs "correlative" approaches (aimed at making
 predictions via applying formal statistical procedures without the
 need to understand molecular mechanisms). In this connection I am
 _very_ interested in collecting a whole spectrum of current opinions,
 approaches and ideas on this topic. Any suggestion or reference will
 be welcome and appreciated. All responses shall be summarized.
 --
 -------------------------------------------------------------------------
 Dr. Igor I. Baskin               !  e-mail: baskin "at@at" qsar.chem.msu.su
 Institute of Organic Chemistry & !  phone : 7-095-939-3557
 Chemistry Department,            !  FAX   : 7-095-939-0290
 Moscow State University,         !
 Moscow, Russia                   !
 --------------------------------------------------------------------------