Local neighborhood behavior in a combinatorial library context.

Horvath, Dragos; Koch, Christian; Schneider, Gisbert; Marcou, Gilles; Varnek, Alexandre

This article revisits a particular aspect of the molecular similarity principle-the Neighborhood Behavior (NB) concept. Earlier, the NB optimality criterion was introduced to select descriptor spaces, combining a given descriptor set and a similarity metric, which optimally comply with the similarity principle. Here, we focus on a 'local' analysis based on the neighborhood of individual bioactive compounds. The defined NB-score measures similarity-based virtual screening success when using individual actives as queries. Systematic studies of local NB have been performed on a large combinatorial library of compounds with reported IC values for five proteases, involving more than 140 descriptor/metric combinations of various fragment- and pharmacophore-based descriptors and different similarity metrics. Although, for each descriptor/metrics combination, the NB-score heavily depends on the query compound, on the average, 2D pharmacophore-based descriptors outperformed their 3D counterparts.

PROTEOLYTIC enzymes; PROTEASE inhibitors; DIAGNOSTIC imaging; BIOACTIVE compounds; MEDICAL screening; HOMOGENEITY

Journal of Computer-Aided Molecular Design, 2011, Vol 25, Issue 3, p237

0920-654X

Academic Journal

10.1007/s10822-011-9416-2