Similarity-based descriptors (SIBAR) as tool for exchange of chemical information

COMP 47

Gerhard F. Ecker, gerhard.f.ecker@univie.ac.at1, Barbara Zdrazil2, and Dominik Kaiser, dominik.kaiser@univie.ac.at1. (1) Department of Pharmaceutical Chemistry, University of Vienna, Althanstrasse 14, Vienna, A-1090, Austria, (2) Department of Medicinal Chemistry, University of Vienna, Althanstrasse 14, Vienna, A-1090, Austria
Recently we published the successful application of a set of new descriptors based on similarity values, denoted as SIBAR-descriptors (Similarity Based SAR). These descriptors are based on calculation of similarity (on basis of euclidian distances) for each compound of the data set to each compound of a reference set, using common descriptors. These euclidian distances (= similarity values) are then further used for QSAR-studies. Both the references set as well as the descriptors used for calculating the SIBAR-values are tailored to the specific QSAR-problem. Best results have been obtained when targeting ADMET-problems. In any case it needs the knowledge of the reference set to retrieve the corresponding descriptors. Assuming that only the descriptors for calculating the SIBAR-values, but not the structures of the reference compounds are available, it should be impossible to trace back the chemical structure of the original compounds of the training set.