Similarity Searching in Files of Three-Dimensional Chemical Structures:  Evaluation of the EVA Descriptor and Combination of Rankings Using Data Fusion

Abstract

EVA is a new molecular descriptor that provides a concise summary of the fundamental frequency components of a molecule's infrared range vibrational spectrum in a vector format. Target structures from the Starlist database are used to demonstrate the effectiveness of the descriptor for similarity searching and its difference from a conventional similarity measure based on the matching of two-dimensional (2D) fingerprints. The use of data fusion on the rankings resulting from the EVA-based and the 2D-based similarity measures results in a combined ranking that can be more effective in simulated property prediction experiments than either of the individual rankings.