CLUES Compound Validation — Reference Databases (HMDB & CPDat)

This Zenodo record hosts the two prebuilt SQLite reference databases used by the clues-confirmed-compound-validation workflow for chemical-source classification (endogenous vs. exogenous, Level 1 / Level 2). They are distributed here because they are too large to commit to the software repository. Both files are optional for the workflow — without them it falls back to public web APIs (slower, slightly lower coverage). They exist to make classification fast, offline, and reproducible. Files in this record File Size SHA-256 Table Rows hmdb_v5.0_metabolites_2026-03.db 623,128,576 B (~594 MB) 384449158a6507bb1d4a8c5cb1f7b14e622f019862d2421ea058439b798ef65f metabolites 217,920 cpdat_v4.0_chemicals_2026-03.db 21,565,440 B (~21 MB) 8674e7398d9e9f76060089a825857f496701312740f5071ca99d8e0aebdd0876 chemicals 33,859 Verify a download with: shasum -a 256 hmdb_v5.0_metabolites_2026-03.db cpdat_v4.0_chemicals_2026-03.db What each database contains hmdb_v5.0_metabolites_2026-03.db — endogenous evidence (from HMDB) SQLite table metabolites (217,920 rows), one row per HMDB metabolite, with columns: hmdb_id, inchikey, name, cas, pubchem_cid, chebi_id, kegg_id, source_tags, classification, description. The workflow looks up compounds by InChIKey-14 (the first 14 characters of the InChIKey) and uses the source_tags, classification, and description fields as endogenous evidence in Chemsource Level 1 (endo/exo) and Level 2 (super-pathway). Provenance: built from the public HMDB "All Metabolites" bulk XML (https://hmdb.ca/downloads) using the build_hmdb_db.py script in the software repo. Snapshot built 2026-03 (HMDB 5.0). Only the fields above are extracted from the full XML; descriptions and the Disposition→Source ontology tags are retained for classification. cpdat_v4.0_chemicals_2026-03.db — exposure context (from EPA CompTox / CPDat) SQLite table chemicals (33,859 rows), one row per DSSTox substance, with columns: dtxsid, inchikey, casrn, preferred_name, functional_use, product_categories, chemical_lists, description. Used for exposure-source context of exogenous compounds in Chemsource Level 2 (functional use, product categories, source lists). Provenance: built from EPA CompTox Chemicals Dashboard / CPDat v4.0 (Figshare) plus the DSSTox SDF dump, using the build_cpdat_db.py script. Snapshot built 2026-03. Not included here (external dependency) The workflow also uses PubChemLite for Exposomics (CCSbase) (~295 MB, ~567k rows, matched by InChIKey-14). This is a third-party dataset already published on Zenodo by the Schymanski group — do not re-upload it. Point the workflow at that existing record instead. Cite it separately (below). How to use these files with the workflow Get the software: https://github.com/CLUES-Emory/clues-confirmed-compound-validation Install the dependencies (see the repository's requirements file). Fetch this data automatically — paste this record's direct-download URLs into the workflow's data_sources.yaml (the hmdb_db and cpdat_db URL fields; the SHA-256 and byte sizes are already prefilled there), then run the setup script: python scripts/setup_data.py This downloads both files into the data/ directory and verifies their checksums. Or place them manually: drop hmdb_v5.0_metabolites_2026-03.db and cpdat_v4.0_chemicals_2026-03.db into the repository's data/ directory. Rebuilding from source These databases can be regenerated from their public upstream sources with the build scripts shipped in the software repository: # HMDB — after downloading "All Metabolites" XML from https://hmdb.ca/downloads python scripts/build_dbs/build_hmdb_db.py --xml hmdb_metabolites.xml.gz --db data/hmdb_v5.0_metabolites_2026-03.db # CPDat — from CPDat v4.0 (Figshare) + DSSTox SDF dump (EPA CompTox) python scripts/build_dbs/build_cpdat_db.py --cpdat-dir CPDAT_DIR --sdf-dir DSSTOX_SDF_DIR --db data/cpdat_v4.0_chemicals_2026-03.db Rebuilding from newer upstream releases will change row counts and checksums. Licensing These are derived databases built from third-party sources; the upstream terms govern reuse of the underlying data: HMDB data is free for academic and non-commercial use with attribution; commercial redistribution requires permission. See https://hmdb.ca/about. This record is released under CC BY-NC 4.0 to remain consistent with HMDB's non-commercial terms. EPA CompTox / DSSTox / CPDat data is U.S. EPA public data (effectively public domain / CC0). The build scripts and the clues-confirmed-compound-validation software are MIT-licensed (separate from these data files). How to cite This data record: (add the Zenodo DOI once minted) Underlying sources — please also cite: HMDB: Wishart DS, et al. HMDB 5.0: the Human Metabolome Database for 2022. Nucleic Acids Research. 2022;50(D1):D622–D631. CPDat: Dionisio KL, et al. The Chemical and Products Database (CPDat), a resource for exposure-relevant data on chemicals in consumer products. Scientific Data. 2018;5:180125. EPA CompTox Dashboard: Williams AJ, et al. The CompTox Chemistry Dashboard: a community data resource for environmental chemistry. J Cheminformatics. 2017;9:61. PubChemLite for Exposomics (if used): cite the PubChemLite for Exposomics Zenodo record (Schymanski group) and its associated publication. Generated for the CLUES (Comprehensive Laboratory for Untargeted Exposome Science) research group, Emory Rollins School of Public Health.