Prediction of biological activity spectra for substances: Evaluation on the diverse sets of drug-like structures

The concept of Biological Activity Spectrum served as a basis for developing PASS (Prediction of Activity Spectra for Substances) software product. PASS predicts simultaneously more than 780 pharmacological effects and biochemical mechanisms based on the structural formula of a substance. It may be used for finding new targets (mechanisms) for known pharmaceuticals and for searching new biologically active substances. PASS prediction ability was evaluated by activity spectra prediction for 63 substances that are presented in the Molecule of the Month section of Prous Science (http://www.prous.com) belong to different chemical classes and reveal various types of biological activity. Mean accuracy of prediction turned out to be about 90%; therefore, it is reasonable to use PASS for finding and optimizing new lead compounds. A web-site with a new internet version of PASS is introduced into practice in December 2001 (http://www,ibmh.msk.su/PASS). On the site, one can find a detailed description of the PASS approach as well as some examples of its applications, and estimate the quality of prediction by submitting structures of substances with known activities.