Database fingerprint (DFP): an approach to represent molecular databases

Abstract

Background: Molecular fingerprints are widely used in several areas of chemoinformatics including diversity analysis and similarity searching. The fingerprint-based analysis of chemical libraries, in particular of large collections, usually requires the molecular representation of each compound in the library that may lead to issues of storage space and redundant calculations. In fact, information redundancy is inherent to the data, resulting on binary digit positions in the fingerprint without significant information.

Results: Herein is proposed a general approach to represent an entire compound library with a single binary fingerprint. The development of the database fingerprint (DFP) is illustrated first using a short fingerprint (MACCS keys) for 10 data sets of general interest in chemistry. The application of the DFP is further shown with PubChem fingerprints for the data sets used in the primary example but with a larger number of compounds, up to 25,000 molecules. The performance of DFP were studied through differential Shannon entropy, k-mean clustering, and DFP/Tanimoto similarity.

Conclusions: The DFP is designed to capture key information of the compound collection and can be used to compare and assess the diversity of molecular libraries. This Preliminary Communication shows the potential of the novel fingerprint to conduct inter-library relationships. A major future goal is to apply the DFP for virtual screening and developing DFP for other data sets based on several different type of fingerprints.Graphical AbstractDatabase fingerprint captures the key information of molecular databases to perform chemical space characterization and virtual screening.

Keywords: Diversity; Information content; Molecular fingerprints; Shannon entropy; Similarity.

Graphical Abstract

Database fingerprint captures the key information of molecular databases to perform chemical space characterization and virtual screening

Fig. 1

a Schematic representation of a binary and dictionary-based molecular fingerprint. b Schematic representation of a database fingerprint (DFP)

Fig. 2

Overview of the approach implemented in this work

Fig. 3

Probability distributions of MACCS keys (166-bits) of representative data sets studied in this work. The number of compounds, mean MACCS keys/Tanimto similarity, and Shannon entropy (SE) are shown

Fig. 4

Relationship Shannon Entropy and MACCS keys/Tanimoto similarity for the ten compound data sets in Table 1. A drugs, I general screening, C clinical, G GDB13, D DNMT1, E epigenetic focused, M semi-synthetic, N natural products, B benzimidazole, GR GRAS, R random

Fig. 5

Relationship Shannon entropy and DFP/Tanimoto similarity and k-mean Euclidean clustering for the ten compound data sets in Table 2 at threshold of 0.5 threshold value

Fig. 6

Probability distribution of the 198 significant bit positions recovered from the original databases represented by PubChem fingerprint at threshold of 0.5