# Cohort-based strategies as an in-house tool to evaluate and improve phenotyping robustness of LC–MS/MS lipidomics platforms

## TL;DR

The paper introduces a method to assess and improve the reliability of lipidomics data using sphingolipid subsets and statistical analysis.

## Contribution

A novel in-house validation methodology is proposed to evaluate lipid reliability and methodological aspects in lipidomics.

## Key findings

- The approach identifies optimal labeled internal standards and concentrations for lipid analysis.
- It enables detection of coelutions with identical monitoring transitions in lipidomics data.
- The method highlights areas for improvement in lipidomics platform robustness.

## Abstract

In recent years, instrumental improvements have enabled the spread of mass spectrometry–based lipidomics platforms in biomedical research. In mass spectrometry, the reliability of generated data varies for each compound, contingent on, among other factors, the availability of labeled internal standards. It is challenging to evaluate the data for lipids without specific labeled internal standards, especially when dozens to hundreds of lipids are measured simultaneously. Thus, evaluation of the performance of these platforms at the individual lipid level in interlaboratory studies is generally not feasible in a time-effective manner. Herein, using a focused subset of sphingolipids, we present an in-house validation methodology for individual lipid reliability assessment, tailored to the statistical analysis to be applied. Moreover, this approach enables the evaluation of various methodological aspects, including discerning coelutions sharing identical selected reaction monitoring transitions, pinpointing optimal labeled internal standards and their concentrations, and evaluating different extraction techniques. While the full validation according to analytical guidelines for all lipids included in a lipidomics method is currently not possible, this process shows areas to focus on for subsequent method development iterations as well as the robustness of data generated across diverse methodologies.

The online version contains supplementary material available at 10.1007/s00216-024-05404-8.