Contribution to the Italian Metabolomics Network General Meeting 2023
This repository contains the code and presentation of my contribution to the Italian Metabolomics Network General Meeting 2023.
Original files of data showed in the presentation are available
on MetaboLights (MTBLS5261),
whereas the source code is in the file
"code.Rmd" of this repository.
Abstract
Boosting annotation confidence in untargeted lipidomics experiments by the use of complementary chemical properties
Liquid chromatography-mass spectrometry is currently the predominant analytical technique in the field of lipidomics. However, even coupling high mass resolution spectrometers with extensive reference libraries, limitations on the annotation phase still persist. In fact, it has been demonstrated that an unambiguous annotation cannot be obtained even relying on m/z and fragmentation experiments (Köfeler et al. Nat Commun. 2021;12:4771). Misannotations can have dramatic consequences in the biological interpretation stage, in particular in systems - as plants or microorganisms - where it is not uncommon to find nonstandard lipid classes. Embedding retention time information in the annotation pipeline represents an important step not only to reduce false annotations, but also to expand the annotation capacity to those features for which it is not possible to get the fragmentation patterns. In this contribution we present an R pipeline which can be used to perform high-confidence accurate lipid identification, combining various types of complementary chemical properties. To illustrate the proposed approach we use the grape lipidome as proof of concept (Garcia-Aloy et al. Food Chem. 2023;410:135360). The application of this workflow to the analytical data allowed a 60% reduction of potential erroneous annotations by considering how each lipid class is ionised, the MS/MS spectras and by the use of retention time dependencies plots. In the contribution, we will also emphasize the usefulness of relying on analytical standards to identify class specific analytical patterns/trends. With the application of the suggested pipeline to the mentioned dataset we also demonstrate how it is possible to spot less explored lipid classes or even new potential ones.