Machine learning–enabled non–targeted metabolomics reveals nutritional and metabolic responses of Brachypodium distachyon to drought and elevated CO2
Data files
Sep 16, 2025 version files 1.03 GB
- 
              
                DesMarais_et_al_JEXBOT2025_NegativeMode_area.txt
                10.12 MB
- 
              
                DesMarais_et_al_JEXBOT2025_NegativeMode_height.txt
                9.91 MB
- 
              
                DesMarais_et_al_JEXBOT2025_NegativeMode_mzML.zip
                592.52 MB
- 
              
                DesMarais_et_al_JEXBOT2025_NegativeMode_sn.txt
                7.47 MB
- 
              
                DesMarais_et_al_JEXBOT2025_PositiveMode_area.txt
                24.71 MB
- 
              
                DesMarais_et_al_JEXBOT2025_PositiveMode_height.txt
                24.36 MB
- 
              
                DesMarais_et_al_JEXBOT2025_PositiveMode_mzML.zip
                338.46 MB
- 
              
                DesMarais_et_al_JEXBOT2025_PositiveMode_sn.txt
                24 MB
- 
              
                README.md
                4.90 KB
Abstract
Rising atmospheric CO2, coupled with intensified drought in many regions, impacts the physiology of C3 plants beyond photosynthesis and carbon metabolism. The interaction between CO2 and drought affects many plant nutrient concentrations, reducing the CO2 fertilization effect in natural systems, increasing agrochemical use, and reducing the nutritive content of crops. To address these challenges, we investigated nutrient dynamics in Brachypodium distachyon, a model for C3 cereal grasses, under ambient (400 ppm) and elevated (800 ppm) CO2, factorially combined with well-watered or drought treatments. Integrative analyses of plant physiology, ionomics, transcriptomics, and non-targeted metabolomics revealed that plant elemental composition and metabolomic responses to elevated CO2 strongly depend on water availability. Elevated CO2 and drought impaired nitrogen status, with root nitrate uptake being more negatively affected than ammonium uptake. However, elevated CO2 increased iron partitioning in leaves under drought, potentially driven by enhanced carbon availability, facilitating chelator synthesis for iron translocation. The high accumulation of sphingolipids in roots under combined stresses suggests a protective role against ionome imbalances. These findings highlight how environmental stressors shape plant nutrient dynamics, providing insights that may improve modeling ecosystem response or guide agricultural practices and breeding strategies to optimize nutrient management under changing climate conditions.
This dataset contains raw LC–MS (.mzML) and processed metabolite feature tables for shoot and root tissues of Brachypodium distachyon exposed to factorial combinations of ambient vs elevated CO2 and control vs drought water regimes. Each .mzML corresponds to a single biological replicate. Processed tables provide integrated peak areas, peak heights, and signal-to-noise ratios. These data were used in multivariate and machine learning analyses described in the accompanying manuscript (J Exp Bot 2025, https://doi.org/10.1093/jxb/eraf382).
Descriptions
Raw data
- 
DesMarais_et_al_JEXBOT2025_PositiveMode_mzML.zip Centroided LC–MS data acquired in positive ionization mode. Each file corresponds to one biological sample. 
- 
DesMarais_et_al_JEXBOT2025_NegativeMode_mzML.zip Centroided LC–MS data acquired in negative ionization mode. 
Each .mzML filename encodes treatment, replicate, and tissue:
- Treatment codes:
- 400CR, 400CS, 400DR, 400DS = 400 ppm CO2; Control or Drought; Root or Shoot
- 800CR, 800CS, 800DR, 800DS = 800 ppm CO2; Control or Drought; Root or Shoot
 
- Replicate: biological replicate number (1, 2, 3)
- Tissue: Root or Shoot
Processed feature tables (tab-delimited .txt)
Positive mode
- DesMarais_et_al_JEXBOT2025_PositiveMode_area.txt — Peak areas
- DesMarais_et_al_JEXBOT2025_PositiveMode_height.txt — Peak heights
- DesMarais_et_al_JEXBOT2025_PositiveMode_sn.txt — Signal-to-noise ratios
Negative mode
- DesMarais_et_al_JEXBOT2025_NegativeMode_area.txt — Peak areas
- DesMarais_et_al_JEXBOT2025_NegativeMode_height.txt — Peak heights
- DesMarais_et_al_JEXBOT2025_NegativeMode_sn.txt — Signal-to-noise ratios
Table structure
Each processed table (area, height, S/N) includes the following columns:
- Alignment ID — unique feature identifier assigned by software
- Average Rt (min) — retention time in minutes
- Average Mz — average m/z of detected feature
- Metabolite name — putative annotation if available
- Adduct type — ion adduct (e.g., [M+H]+, [M-H]−)
- Post curation result — curation status flag
- Fill % — feature filling rate across samples
- MS/MS assigned — whether MS/MS spectrum is assigned
- Reference RT / m/z — reference values for annotation
- Formula — chemical formula (if annotated)
- Ontology — compound class/ontology assignment
- INCHIKEY — InChIKey identifier
- SMILES — SMILES string
- Annotation tag (VS1.0) — annotation confidence tag
- RT matched / m/z matched / MS/MS matched — matching criteria used
- Comment — notes on annotation
- Manually modified for quantification / annotation — user curation flags
- Isotope tracking parent ID/weight number — isotope-tracking metadata
- Total score — overall annotation confidence score
- RT similarity / Dot product / Reverse dot product / Fragment presence % — spectral matching metrics
- S/N average — signal-to-noise ratio
- Spectrum reference file name — reference spectrum file
- MS1 isotopic spectrum / MS/MS spectrum — spectral data references
Following these metadata columns, each table contains sample-level intensity values as additional columns (sample IDs follow the .mzML naming scheme, e.g., 400CR_1, 800DS_3).
Code / Software
- Raw file conversion: ProteoWizard msConvert (centroided).
- Feature detection and quantification: MS-DIAL v4.9.
- All exported tables were used directly in multivariate and machine learning analyses performed in R (v4.3.1) and Python (v3.10).
No scripts are included in this Dryad submission. Processing workflows are described in the manuscript and Supporting Information.
File-level descriptions
DesMarais_et_al_JEXBOT2025_PositiveMode_mzML.zip — Centroided raw LC–MS files, positive ionization.
DesMarais_et_al_JEXBOT2025_PositiveMode_area.txt — Feature table with peak area values.
DesMarais_et_al_JEXBOT2025_PositiveMode_height.txt — Peak heights.
DesMarais_et_al_JEXBOT2025_PositiveMode_sn.txt — Signal-to-noise ratios.
DesMarais_et_al_JEXBOT2025_NegativeMode_mzML.zip — Centroided raw LC–MS files, negative ionization.
DesMarais_et_al_JEXBOT2025_NegativeMode_area.txt — Peak areas (filename typo retained).
DesMarais_et_al_JEXBOT2025_NegativeMode_height.txt — Peak heights.
DesMarais_et_al_JEXBOT2025_NegativeMode_sn.txt — Signal-to-noise ratios.
Contact
- Hsin-Fang Chang — hfchang@mit.edu
- David Des Marais — dldesmar@mit.edu
