Method of predicting biological activity profiles based on the neighborhood behavior of 'in silico' similarity metrics calibrated with respect to activity spaces

摘要

The present invention uses the analysis of the experimental responses of compounds with respect to a panel of multiple activity tests to optimize the Neighborhood Behavior of molecular similarity metrics in a Structural Space (the assumption that structurally similar compounds display similar biological properties), and subsequently applies the herein established relationship between the structural space and the activity test panel in order to predict the most likely activity profiles of a novel compound on the basis of the (experimentally measured) ones of its nearest neighbors among a reference set of drugs, druglike molecule and other chemical compounds. The dual nature of the Neighborhood Behavior is described in terms of complementary aspects following complementary quality criteria to calibrate and define the proficiency of In Silico structural space to map on the Activity Space.