Data Base similarity (DBsimilarity) of Natural Products to aid compound identification on MS and NMR pipelines, similarity networking and more

Abstract

Introduction: This paper proposes DBsimilarity to organize structural databases into Similarity Networks to better understand the rich information available. Method: DBsimilarity was written in Jupyter Notebooks to be easy to follow and values readability. It converts SDF files into CSV files, adds chemoinformatics data, constructs a MZMine custom database file and a NMRfilter candidate list of compounds for rapid dereplication of MS and 2D NMR data, calculates similarities between compounds, and constructs CSV files to be converted to Similarity Networks using Cytoscape. Results: The Lotus database was used as source for Ginkgo biloba compounds and DBsimilarity was used to create Similarity Networks that includes NPClassifier classification to indicate biosynthesis pathways. Following, a database of validated antibiotics natural products was combined with the G. biloba database to indicate promising compounds. The presence of 11 compounds in both datasets points to a possible antibiotic property of G. biloba, and 122 other compounds similar to those known antibiotics is found. Next, DBsimilarity was used to filter the NPAtlas database (selecting only those with MIBIG reference) to identify potential antibacterial compounds using the ChEMBL database as reference. It was possible to promptly identify 5 compounds found in both databases, and 167 other worth investigating compounds similar to those known antibiotics. Conclusion: Chemical and biological properties are determined by molecular structures. DBsimilarity enables the creation of interactive Similarity Networks using Cytoscape. It is also in line with recent review that highlights significant sources of errors in compound identification: poor biological plausibility and unrealistic chromatographic behaviors.