Using KNIME to unite open-source tools for drug metabolite identification from UHPLC-MS/MS data

Abstract

An active and growing eco-system of open-source software for data science is emerging. In the area of metabolomics, this facilitates the analysis of increasingly complex datasets. However, these approaches suffer from computational complexity, high theory levels and the multitude of different file formats making integration and interoperation challenging. This project aims to address this by providing an efficient method for the identification of metabolites and improved accuracy of spectral prediction for each candidate metabolite. To that end, a computational workflow has been developed using the open-source KNIME hosting platform for the identification of metabolites from UHPLC-MS/MS data.
This unique workflow, starting from the input of the molecular structure, predicts metabolites using the metabolomics tool SyGMa and generates in-silico MS/MS spectra using CFM-ID. With these predictions in hand, the UHPLC-MS/MS data can then be used to score metabolites and generate EICs for quantification and data analysis. The computational tool MetFrag, has also been integrated into the workflow, to apply the bond disconnection approach to score candidate structures and to match them to the MS/MS spectrum of interest.
To test the utility of the workflow, the metabolism of the drugs amitriptyline and verapamil, employing three different microsomes types, was studied. The metabolites detected were compared to the documented metabolites in the literature. The results revealed several unpublished metabolic reactions and metabolites for both drugs despite them having been studied for many decades.
This illustrated the strength of the workflow, so the next step was to apply it to the analysis of the metabolism of remdesivir - a relatively unstudied drug. This resulted in the detection of several new metabolites resulting from unpublished metabolic reactions.
This workflow solution is flexible, efficient, accurate and reproducible and it will offer an important opportunity to unite and aid the metabolomics community.

Date of Award	6 Dec 2022
Original language	English
Awarding Institution	University of Bristol
Supervisor	Chris Arthur (Supervisor) & Paul J Gates (Supervisor)