Multi-Laboratory Study Reveals Challenges and Opportunities in Untargeted Mass Spectrometry Metabolomics Annotation

A recent collaborative effort involving 10 laboratories has shed light on the complexities and limitations of current approaches to untargeted metabolomics data annotation using mass spectrometry (MS). The study, published in Analytical Chemistry, systematically compares annotation strategies across different platforms and teams, providing valuable insights aimed at honing the future of metabolomics research (1).

Untargeted metabolomics, often powered by liquid chromatography–mass spectrometry (LC–MS), is a cornerstone of systems biology, biomarker discovery, and exposomics. However, rapid identification of known analytes in these samples is a bottleneck that needs to be addressed. This study focused on analyzing an extract from Withania somniferaL. (ashwagandha), a plant used for centuries in traditional Indian medicine, using LC–MS. Multiple datasets—generated via orbital ion trap and quadrupole time-of-flight (QTOF) platforms—were distributed among 10 teams. Each team independently annotated the datasets without prior access to the reference standards, simulating an untargeted approach typical in metabolomics workflows. The primary goal was to assess the consistency and accuracy of annotations and to identify bottlenecks in the process.

The analysis revealed a complex landscape of annotation performance. Across all datasets, the teams collectively identified 142 analytes at the putative level. However, individual teams only detected between 24% and 57% of these, highlighting significant variability. The overlap between teams was higher for feature detection but diminished considerably at the levels of ion species, chemical class, and definitive identity, indicating challenges in consistent annotation.

One issue identified was the occurrence of false positives arising from in-source fragmentation and the formation of redundant features. Many detected features resulted from different adducts, fragment ions, or in-source clusters, which, if not properly annotated, can inflate the perceived complexity of the sample and hinder accurate identification. The study underscores the importance of careful data preprocessing and feature grouping to mitigate these effects.

In addition, the scarcity of overlapping spectral data in open-access repositories posed a major obstacle. Many compounds, especially plant secondary metabolites like withanolides, lack comprehensive MS/MS spectra in open-access databases. The authors emphasize that current annotation pipelines often overestimate the true complexity of the data. The temptation to interpret a high number of detected features as corresponding to unique analytes may lead to overestimations of sample diversity. The study advocates for improved annotation strategies that incorporate multiple lines of evidence, including retention time prediction, in silico fragmentation, and literature verification, alongside spectral matching.

These results serve to highlight the value of collaborative, multi-team approaches. Combining annotations from various pipelines enhances confidence and can generate a more comprehensive picture of the metabolome. This consensus-building process demonstrates that leveraging different strengths and databases across teams could improve overall annotation quality.

While the detection of a vast array of features is feasible, accurate and consistent annotation remains a significant challenge. As the field progresses, fostering open data sharing, standardization, and collaborative validation will all help to transform the promise of untargeted mass spectrometry metabolomics into routine, reliable applications for biological and chemical insights.

Reference

(1) Houriet, J.; Cech, N. B.; et al. Multilaboratory Untargeted Mass Spectrometry Metabolomics Collaboration to Identify Bottlenecks and Propose Pathways Forward. Anal. Chem. 2025, 97, 16110–16122. DOI: 10.1021/acs.analchem.4c05577