ISSN print edition: 0366-6352 ISSN electronic edition: 1336-9075 Registr. No.: MK SR 9/7 Published monthly

LC-HRMS-untargeted metabolomics for identification and authentication of celery (Apium graveolens) from two related plants with similar morphologies

Relita Florentika, Wulan Tri Wahyuni, and Mohamad Rafi

Department of Chemistry, Faculty of Mathematics and Natural Sciences, IPB University, Bogor, Indonesia

E-mail: mra@apps.ipb.ac.id

Received: 11 July 2025  Accepted: 1 May 2026

Abstract:

Celery shares morphological traits with other Apiaceae species, such as cilantro and parsley, all of which are widely used in commercial products. Once processed into powders or extracts, differentiating these species becomes difficult, increasing the risk of adulteration and compromised efficacy. This study aims to develop an untargeted metabolomics using liquid chromatography-high-resolution mass spectrometry (LC-HRMS) to identify and authenticate celery. The validation of LC-HRMS untargeted metabolomics, paired with partial least squares discriminant analysis (PLS-DA) and principal component analysis (PCA), was expected to resolve the issue of celery adulteration. Metabolite profiles were obtained from samples that were extracted using ultrasonic waves. Optimization of LC-HRMS enabled the separation of apigenin and the detection of 577 metabolites in positive mode. Using the mass-to-ion-charge ratio and peak-intensity data, we applied PCA to reduce the large dataset and PLS-DA to identify potential markers based on variable importance in projection and coefficient scores, which were found only in one species. Six celery markers were senkyunolide A, sedanolide, senkyunolide F, apigenin-O-dihexosyl deoxyhexoside, quillaic acid, and luteolin. The PLS-DA model validation was performed using fivefold cross-validation and a permutation test. The R2 and Q2 values of 0.986 and 0.972, respectively, indicate that the model exhibits high predictive ability, achieving 100% accuracy. Moreover, the p-value < 0.05 indicates that the relationship between the X and Y variables is not coincidental. PCA clearly separated the species, with QC samples clustering tightly. The total of 65.0% of the two PC scores helps prevent overfitting.

Keywords: Celery; Chemometrics; Cilantro; LC-HRMS; Metabolomics; Parsley

Full paper is available at www.springerlink.com.

DOI: 10.1007/s11696-026-05004-z

Chemical Papers 80 (9) 10259–10275 (2026)