For LC-HRMS:
Data were uploaded and processed on the Galaxy Workflow4Metabolomics platform (W4M) (Giacomoni et al., 2015), which permits ion extraction, alignment across samples and peak grouping. Briefly, the workflow included a “CentWave” peak picking, “PeakDensity” for peak grouping, “LOESS/non-linear” and “PeakGroups” for retention time correction, and “CAMERA” for the annotation of isotope peaks, adducts and fragments. Extraction blanks were used as a reference for noise estimation to remove noise and artifact signals from the analysed samples. Blank intensities were compared to sample intensities, and a fold change (FC) was calculated using the “Intensity Check” tool in W4M. Features with a FC < 3.0 were excluded, as they were assumed to be analytical contamination. A signal drift correction was subsequently applied using the “Batch correction” tool, employing a LOESS regression model with a span of 0.8.
Finally, features were filtered using the “Quality Metrics” and “Generic filter” tools. Features with a coefficient of variation (CV) in the QC samples exceeding 30% and those with a (pool CV/sample CV) ratio above 1 were removed.
For LC-HRMS/MS:
The converted data were processed with MZmine (Version 4.4.0) (Schmid et al., 2023). The data treatment workflow started with the “Mass Detection” module to identify ions in MS1 and MS2 scans. Chromatographic peaks were then constructed using the “Chromatogram Builder” module, followed by “Smoothing” to refine peak shapes. Feature detection and deconvolution of overlapping or co-eluting peaks were enhanced using the “Local Minimum Feature Resolver.”
Next, the “13C Isotope Filter” and “Isotopic Peaks Finder” were used to identify isotopic patterns and filter features accordingly. The “Join Aligner” module aligned features across samples. To further refine the dataset, the “Feature List Rows Filter” was applied to remove features not meeting user-defined criteria. The “Peak Finder” was used to handle missing values. Misaligned features were removed with the “Duplicate Peak Filter”, and relationships between features, based on retention time and intensity correlations, were identified using the “Correlation Grouping” module.
The “Ion Identity Networking” step annotated grouped features as ion adducts, in source fragments or multimers. The “Feature List Blank Subtraction” module removed features detected in blank injections with a FC < 200%, ensuring the analysis focused on sample-relevant ions.
"],"repository":["MetaboLights"],"study_status":["Public"],"ptm_modification":[""],"instrument_platform":["Liquid Chromatography MS - positive - reverse-phase"],"chromatography_protocol":["Samples were analysed on a Thermo Scientific UltiMate 3000 UHPLC system coupled to a QExactive Orbitrap mass spectrometer.
The UHPLC injection volume was 10 μL, and separation was achieved on a C18-PFP column (150 x 2.1 mm, 2 μm particle diameter, ACE, supplied by AIT France). The mobile phase consisted of water (A) and ACN (B), both acidified with 0.1% FA, and MeOH (C), flowing at 0.4 mL/min. The gradient started with 100% A, transitioned to 100% B over 10 min, and was held for 6 min before switching to 100% C for system rinsing in 1 min. It was maintained for 5 min, then returned to 100% A within 1 min, followed by a 3 min equilibration, resulting in a total run time of 26 min.
"],"publication":["Decoding Food Chemical Reactivity under Thermal Processing and Formulation Changes by Untargeted LC-HRMS/MS and Feature-Based Molecular Networking."],"submitter_affiliation":["AgroParisTech","AgroParisTech - Université Paris Saclay"],"submitter_name":["Mathieu Cladiere","Soha Farah"],"organism_part":["QC_Pooled","W200_50","W200_15","Blank_Equilibrium","QC_Equilibrium","W200_25","Blank_Extraction","W200_5"],"technology_type":["mass spectrometry assay"],"disease":[""],"extraction_protocol":["Samples were extracted in duplicate (Prep-1 and Prep-2) for each condition, and each resulting extract was injected in duplicate. Two extraction blanks, prepared by omitting the sponge cake matrix in the first step, were extracted identically. In addition, pooled quality control (QC) samples were prepared by combining 0.5 mL aliquots of the final clarified supernatant from each of the 8 extractions. Aliquots of 0.5 mL of each sample, blank, and QC were filtered through 0.2 μm syringeless filters (Mini-Uniprep G2, Whatman) prior to injection.
"],"organism":["QC_Pooled","Blank_Equilibrium","QC_Equilibrium","Blank_Extraction","Sponge cake"],"full_dataset_link":["https://www.ebi.ac.uk/metabolights/MTBLS13919"],"author":["Even Le Roux. SayFood - Food and Bioproduct Engineering.","Valérie Camel. SayFood - Food and Bioproduct Engineering.","Mathieu Cladière. SayFood - Food and Bioproduct Engineering. 22 place de l'agronomie, Palaiseau, 91120, France. mathieu.cladiere@agroparistech.fr.","Mélina Ramos. SayFood - Food and Bioproduct Engineering.","Barbara Rega. SayFood - Food and Bioproduct Engineering.","Soha Farah. SayFood - Food and Bioproduct Engineering. 22 place de l'agronomie, Palaiseau, 91120, France. soha.farah@agroparistech.fr."],"data_transformation_protocol":["The ProteoWizard MSConvertGUI software (Chambers et al., 2012) (Version: 3.0.22259, 64-bit) was used to convert raw data into .mzML files prior to any data treatment.
"],"study_factor":["SampleType"],"submitter_email":["mathieu.cladiere@agroparistech.fr","soha_farah@hotmail.com"],"sample_collection_protocol":["Wheat sponge cakes were prepared as follows: Eggs (45% w/w) and sugar (25% w/w) were beaten together for 10 min. Wheat flour was then gradually incorporated over 1.5 min. After mixing for 30 s, the sunflower oil (5% w/w) was incorporated within 15 s and the batter was beaten for an additional 1 min. This process yielded a batter of 630 g, from which 21 identical cakes were produced by filling 25 g of batter into aluminium moulds (8.0 cm x 4.5 cm x 3.5 cm). Cakes were baked at 200°C for 25 minutes and collected at four baking time points: 5, 15, 25, and 50 minutes. For each time point, 5 cakes were sampled, ground and freeze-dried for 48 hours prior to extraction, in order to obtain a composite sample.
"],"omics_type":["Metabolomics"],"study_design":["ultra-performance liquid chromatography-mass spectrometry","W200_15","untargeted analysis","Blank_Equilibrium","kinetic study","W200_25","Blank_Extraction","feature-based molecular network","QC_Pooled","W200_50","processed food","Thermo Scientific Q Exactive","Maillard Reaction","Thermo Scientific Dionex Ultimate 3000 UHPLC system","QC_Equilibrium","W200_5","Sponge cake"],"curator_keywords":["ultra-performance liquid chromatography-mass spectrometry","W200_15","untargeted analysis","Blank_Equilibrium","kinetic study","W200_25","Blank_Extraction","feature-based molecular network","QC_Pooled","W200_50","processed food","Thermo Scientific Q Exactive","Maillard Reaction","Thermo Scientific Dionex Ultimate 3000 UHPLC system","QC_Equilibrium","W200_5","Sponge cake"],"mass_spectrometry_protocol":["For LC-HRMS:
The QExactive operated in centroid mode (resolution of 70,000 FWHM at m/z 200, m/z range: 60-800) with a heated electrospray ionisation source (HESI) in positive mode. For the Orbitrap MS method, the “AGC target” and “Maximum IT” parameters were set to 1 x 10^6 and 200 ms, respectively. The analytical sequence started with 10 mobile phase blank injections and 10 QC spare aliquots to equilibrate the instrument. Sponge cake samples and extraction blank injections were randomised, and 2 QC samples were injected every 4 to 5 samples.
For LC-dda-HRMS/MS:
The untargeted analysis was conducted in Full full MS/dd-MS2 acquisition mode, combining a full scan MS with a set of data-dependent MS2 scans. The top N parameter was set to 5, meaning the 5 most intense MS1 ions were selected and subsequently fragmented to acquire the MS/MS spectra, with a resolution of 35,000. The analytical sequence started with 10 mobile phase blank injections and 10 QC spare aliquots for instrument equilibration. Sponge cake samples and extraction blank injections were randomised, and 2 QC samples were injected every 4 to 5 samples.
"],"metabolite_name":["HMF"],"additional_accession":[]},"is_claimable":false,"name":"Decoding Food Chemical Reactivity under Thermal Processing Changes by Untargeted LC-HRMS/MS and Feature-Based Molecular Networking","description":"Method applicability, which involves investigating the impact of processing through a kinetic study, in which molecular networks are compared across different baking times to capture the progression of thermally induced reactions.
","dates":{"publication":"2026-09-18","submission":"2026-02-18"},"accession":"MTBLS13919","cross_references":{"MetaboLights":["MTBLC412516"],"ChEBI":["CHEBI:412516"]}}