Mapping the origins of molecular signals in food by integrative analysis of untargeted metabolomics with chemical databases

Alejandro Mendoza Cantu, Julia M. Gauglitz, Wout Bittremieux

Integrative analysis of the PTFI dataset, linking ~24,000 molecular features across 500 foods to curated chemical databases.

Overview

Analysis of the first 500 foods from the Periodic Table of Food Initiative discovery database using external chemical databases to contextualize the findings.

Notebooks

• database_matching.ipynb - Matching PTFI features to external databases (DrugBank, FCCdb, PubChem)
• natural_product_matching.ipynb - Matching to natural product databases (LOTUS, COCONUT, SuperNatural 3)
• non_natural_formulas.ipynb - Identifying non-natural product formulas

Analysis workflow (executed in PTFI environment):

1. ptfi_data_prep.html - Data processing and transformation to long format
2. ptfi_external_db_matches.html - External database matching and visualization (DrugBank, FCCdb, Agrochemical PubChem)
3. ptfi_therapeutic_categories.html - DrugBank therapeutic categories analysis
4. ptfi_f_analysis.html - Fluorinated xenobiotics analysis
5. ptfi_producers.html - Putative Novel producer discoveries

Data

PTFI dataset: 500 foods, 24,721 features (900 annotated, 23,821 formula-only)

Access: https://pmp.heart.org/ (PTFI online platform)

Note: Raw PTFI data is only accessible through the PTFI platform and is not included in this repository.

Database requirements

This analysis requires external chemical databases (not included):

• LOTUS: lotus.naturalproducts.net
• COCONUT: coconut.naturalproducts.net
• DrugBank: drugbank.ca (requires registration)
• SuperNatural 3: Natural products database
• PubChem: Agrochemicals subset

Requirements

Python 3.12 with:

pandas==2.2.3
numpy
matplotlib==3.9.3
seaborn==0.13.2
scipy
pymongo
lxml
jupyter

Key findings

• Contextualization of the PTFI dataset through external dataset matching
• Distinct xenobiotic fingerprints across food categories
• Dietary transfer of bioactive compounds
• Putative novel producer discoveries for bioactive natural products

Citation

Mendoza Cantu, A., Gauglitz, J.M., Bittremieux, W. (2025). 
Mapping the origins of molecular signals in food by integrative 
analysis of untargeted metabolomics with chemical databases. 
[Manuscript in preparation]

Acknowledgments

PTFI 2024 Data Challenge | Research Foundation – Flanders (FWO G0AGQ24N)

Contact: wout.bittremieux@uantwerpen.be