Bayesian inference network for molecular similarity searching using 2D fingerprints and multiple reference structures

Abstract

2D fingerprint based similarity searching using a single bioactive reference is the most popular and effective virtual screening tool. In our last paper, we have introduced a novel method for similarity searching using Bayesian inference network (BIN). In this study, we have compared BIN with other similarity searching methods when multiple bioactive reference molecules are available. Three different 2D fingerprints were used in combination with data fusion and nearest neighbor approaches as search tools and also as descriptors for BIN. Our empirical results show that the BIN consistently outperformed all conventional approaches such as data fusion and nearest neighbor, regardless of the fingeyrints that were tested.