HMDB Database

Overview

The Human Metabolome Database (HMDB) is a comprehensive, freely available resource containing detailed information about small molecule metabolites found in the human body.

Scripts

scripts/query_hmdb.py — fetch and parse an HMDB metabolite XML record by accession (stdlib only, JSON to stdout):

python scripts/query_hmdb.py HMDB0000001    # full accession
python scripts/query_hmdb.py 1              # bare number (zero-padded automatically)

Note: HMDB serves no public REST API and may rate-limit/block automated fetches; for bulk work download the XML/SDF dumps from https://www.hmdb.ca/downloads.

When to Use This Skill

This skill should be used when performing metabolomics research, clinical chemistry, biomarker discovery, or metabolite identification tasks.

Database Contents

HMDB version 5.0 (released 2022; latest major version as of mid-2026) contains:

• 220,945 metabolite entries covering both water-soluble and lipid-soluble compounds (the v5.0 paper reported 217,920; the live site count grows with curation)
• ~8,600 protein sequences for enzymes and transporters involved in metabolism
• 130+ data fields per metabolite including:
  • Chemical properties (structure, formula, molecular weight, InChI, SMILES)
  • Clinical data (biomarker associations, diseases, normal/abnormal concentrations)
  • Biological information (pathways, reactions, locations)
  • Spectroscopic data (NMR, MS, MS-MS spectra)
  • External database links (KEGG, PubChem, MetaCyc, ChEBI, PDB, UniProt, GenBank)

Core Capabilities

1. Web-Based Metabolite Searches

Access HMDB through the web interface at https://www.hmdb.ca/ for:

Text Searches:

• Search by metabolite name, synonym, or identifier (HMDB ID)
• Example HMDB IDs: HMDB0000001, HMDB0001234
• Search by disease associations or pathway involvement
• Query by biological specimen type (urine, serum, CSF, saliva, feces, sweat)

Structure-Based Searches:

• Use ChemQuery for structure and substructure searches
• Search by molecular weight or molecular weight range
• Use SMILES or InChI strings to find compounds

Spectral Searches:

• LC-MS spectral matching
• GC-MS spectral matching
• NMR spectral searches for metabolite identification

Advanced Searches:

• Combine multiple criteria (name, properties, concentration ranges)
• Filter by biological locations or specimen types
• Search by protein/enzyme associations

2. Accessing Metabolite Information

When retrieving metabolite data, HMDB provides:

Chemical Information:

• Systematic name, traditional names, and synonyms
• Chemical formula and molecular weight
• Structure representations (2D/3D, SMILES, InChI, MOL file)
• Chemical taxonomy and classification

Biological Context:

• Metabolic pathways and reactions
• Associated enzymes and transporters
• Subcellular locations
• Biological roles and functions

Clinical Relevance:

• Normal concentration ranges in biological fluids
• Biomarker associations with diseases
• Clinical significance
• Toxicity information when applicable

Analytical Data:

• Experimental and predicted NMR spectra
• MS and MS-MS spectra
• Retention times and chromatographic data
• Reference peaks for identification

3. Downloadable Datasets

HMDB offers bulk data downloads at https://www.hmdb.ca/downloads in multiple formats:

Available Formats:

• XML: Complete metabolite, protein, and spectra data
• SDF: Metabolite structure files for cheminformatics
• FASTA: Protein and gene sequences
• TXT: Raw spectra peak lists
• CSV/TSV: Tabular data exports

Dataset Categories:

• All metabolites or filtered by specimen type
• Protein/enzyme sequences
• Experimental and predicted spectra (NMR, GC-MS, MS-MS)
• Pathway information

Best Practices:

• Download XML format for comprehensive data including all fields
• Use SDF format for structure-based analysis and cheminformatics workflows
• Parse CSV/TSV formats for integration with data analysis pipelines
• Check version dates to ensure up-to-date data (current major version: v5.0)

Usage Requirements:

• Free for academic and non-commercial research
• Commercial use requires explicit permission (contact samackay@ualberta.ca)
• Cite HMDB publication when using data

4. Programmatic Access

HMDB publishes no documented public REST API. Practical programmatic routes, in order of preference:

• Bulk downloads (preferred for any volume): Parse the XML/SDF/CSV dumps from https://www.hmdb.ca/downloads locally. This is the only route that scales and won't get rate-limited.
• Per-record XML endpoint: Each entry is served as XML at https://www.hmdb.ca/metabolites/<ID>.xml (used by scripts/query_hmdb.py). Undocumented and aggressively rate-limited/blocked (HTTP 403/429) for automated clients — fine for a handful of ad-hoc lookups, not for batch jobs. Send a descriptive User-Agent and back off on failures.
• R/Bioconductor hmdbQuery: BiocManager::install("hmdbQuery") wraps the same web endpoints for R workflows.
• Custom API: For sanctioned bulk/commercial API access, contact the HMDB team (see Usage Requirements above for the listed address).

5. Common Research Workflows

Metabolite Identification in Untargeted Metabolomics:

1. Obtain experimental MS or NMR spectra from samples
2. Use HMDB spectral search tools to match against reference spectra
3. Verify candidates by checking molecular weight, retention time, and MS-MS fragmentation
4. Review biological plausibility (expected in specimen type, known pathways)

Biomarker Discovery:

1. Search HMDB for metabolites associated with disease of interest
2. Review concentration ranges in normal vs. disease states
3. Identify metabolites with strong differential abundance
4. Examine pathway context and biological mechanisms
5. Cross-reference with literature via PubMed links

Pathway Analysis:

1. Identify metabolites of interest from experimental data
2. Look up HMDB entries for each metabolite
3. Extract pathway associations and enzymatic reactions
4. Use linked SMPDB (Small Molecule Pathway Database) for pathway diagrams
5. Identify pathway enrichment for biological interpretation

Database Integration:

1. Download HMDB data in XML or CSV format
2. Parse and extract relevant fields for local database
3. Link with external IDs (KEGG, PubChem, ChEBI) for cross-database queries
4. Build local tools or pipelines incorporating HMDB reference data

Related HMDB Resources

The HMDB ecosystem includes related databases:

• DrugBank: ~2,832 drug compounds with pharmaceutical information
• T3DB (Toxin and Toxin Target Database): ~3,670 toxic compounds
• SMPDB (Small Molecule Pathway Database): Pathway diagrams and maps
• FooDB: ~70,000 food component compounds

These databases share similar structure and identifiers, enabling integrated queries across human metabolome, drug, toxin, and food databases.

Best Practices

Data Quality:

• Verify metabolite identifications with multiple evidence types (spectra, structure, properties)
• Check experimental vs. predicted data quality indicators
• Review citations and evidence for biomarker associations

Version Tracking:

• Note HMDB version used in research (current: v5.0)
• Databases are updated periodically with new entries and corrections
• Re-query for updates when publishing to ensure current information

Citation:

• Always cite HMDB in publications using the database
• Reference specific HMDB IDs when discussing metabolites
• Acknowledge data sources for downloaded datasets

Performance:

• For large-scale analysis, download complete datasets rather than repeated web queries
• Use appropriate file formats (XML for comprehensive data, CSV for tabular analysis)
• Consider local caching of frequently accessed metabolite information

Reference Documentation

See references/hmdb_data_fields.md for detailed information about available data fields and their meanings.