Bayesian inference network significantly improves the effectiveness of similarity searching using multiple 2D fingerprints and multiple reference structures

Abstract

Recent work in similarity searching have suggested that significant improvements in retrieval effectiveness can be achieved by combining results from multiple reference structures or multiple molecular descriptors. Recently, the Bayesian inference network model (BIN) has been introduced for performing molecular similarity searching using a single and multiple reference structures. One of the important characteristics of the inference network model is that it permits the combination of multiple reference structures and multiple molecular descriptors. This paper introduces an inference network model developed for molecular similarity searching that integrates into a single framework, multiple reference structures and multiple molecular descriptors. The inference network model of similarity, which was designed from this point of view, treats similarity searching as an evidential reasoning process where multiple sources of evidence about reference and compound content are combined to estimate similarity scores. Our results show that, the BIN with multiple descriptors is notably more effective than BIN with a single descriptor in searches for structurally diverse sets of actives molecules.