Communication Dans Un Congrès Année : 2026

Does Data Conversion Matter ? Assessing Its Impact on LC–MS Untargeted Metabolomics

La conversion des données est-elle importante ? Évaluation de son impact sur la métabolomique non ciblée LC-MS

Thomas Alexandre Brunet
Pierre Lanteri
Yohann Clément

Résumé

Untargeted LC–MS metabolomics generates highly complex datasets, making chemometric methods essential for exploring and interpreting the data. Although much attention has been given to signal detection, normalization, and statistical modeling, the initial step of data conversion — especially the transformation from profile to centroid mode — is rarely examined, even though it directly influences variance structure and relationships between variables. In this study, biological samples were analyzed in profile mode and then converted into centroided data using three common approaches: a wavelet‑based method, a vendor‑embedded algorithm, and a proprietary converter. Chemometric analyses applied to these converted datasets reveal that each centroiding strategy significantly alters the multivariate structure. Differences were observed in point density, intensity distributions, and the amount of retained information. These discrepancies affected PCA and t‑SNE representations, where samples grouped more according to the conversion method than their biological origin. Supervised models also showed that the discriminant regions of the dataset varied depending on the algorithm used, potentially leading to biased interpretations. Overall, the results demonstrate that data conversion is not a neutral preprocessing step. It strongly shapes the statistical structure of LC–MS data and can introduce hidden variability. Therefore, centroiding choices must be explicitly considered within chemometric validation workflows to improve the robustness and interpretability of metabolomics analyses.

Fichier non déposé

Dates et versions

hal-05522522 , version 1 (22-02-2026)

Identifiants

  • HAL Id : hal-05522522 , version 1

Citer

Rémy de Boni, Arnaud Salvador, Thomas Alexandre Brunet, Valentina Calabrese, Delphine Arquier, et al.. Does Data Conversion Matter ? Assessing Its Impact on LC–MS Untargeted Metabolomics. CHEMIOMETRIE XXV, Groupe Français de Chimiométrie, Feb 2026, Nancy, France. ⟨hal-05522522⟩
82 Consultations
0 Téléchargements

Partager

  • More