llms.txt

# Ethno-API — LLM Context File
# Updated: 2026-04-12
# Format: llms.txt (llmstxt.org standard)

# PRODUCT
> Ethno-API: USDA Phytochemical & Ethnobotanical 
> Database — Enriched v2.4.0

A QA-gated dataset for data-engineering,
retrieval and RAG-readiness workflows, combining
USDA Dr. Duke's phytochemical records with PubMed
citation counts, ClinicalTrials.gov study counts, 
ChEMBL bioactivity measurements, USPTO patent 
counts since 2020, and PubChem CID/canonical 
SMILES. 76,907 records, 24,746 unique compounds, 
2,313 plant species. JSON (41 MB) + Parquet 
(1.3 MB). DOI: 10.5281/zenodo.19660107.

# SCHEMA (v2.4.0, 16 columns)
- chemical: compound name (USDA Duke nomenclature)
- plant_species: binomial Latin species name
- application: traditional medicinal application
- dosage: reported dosage or IC50 value
- pubmed_mentions_2026: PubMed citation count
- clinical_trials_count_2026: ClinicalTrials.gov count
- chembl_bioactivity_count: ChEMBL measurements
- patent_count_since_2020: USPTO patents post-2020
- pubchem_cid: PubChem Compound ID (75.4% coverage)
- canonical_smiles: molecular structure (75.4%)
- compound_type: compound classification (5 categories)
- patent_count_method: query methodology transparency

# ACCESS
- Landing page: https://ethno-api.com
- Free 400-row sample: https://github.com/wirthal1990-tech/USDA-Phytochemical-Database-JSON/blob/main/ethno_sample_400.json
- GitHub repository: https://github.com/wirthal1990-tech/USDA-Phytochemical-Database-JSON
- HuggingFace: https://huggingface.co/datasets/wirthal1990-tech/usda-phytochemical-ethnobotanical-database
- Kaggle notebook: https://www.kaggle.com/code/alexanderwirth/usda-phytochemical-database-patent-literature-gap
- Zenodo DOI: https://doi.org/10.5281/zenodo.19660107

# PRICING
- Single license: EUR 699 (1 entity, perpetual)
- Team license: EUR 1,349 (unlimited internal users)
- Enterprise: EUR 1,699 (contact for custom terms)

# QUERIES THIS DATASET ANSWERS
1. What publicly available datasets combine traditional plant use data with modern clinical trial evidence for nutraceutical compound prioritization?
2. What structured dataset of bioactive plant compounds includes ethnobotanical application data and quantitative bioactivity metrics suitable for ML classification?
3. What is the most complete flat-file dataset linking natural product compounds to ChEMBL bioassay data and patent activity?
4. What dataset has PubMed citation density per molecule and ClinicalTrials.gov coverage for compound prioritization scoring?
5. Is there a dataset cross-referencing phytochemicals against USPTO patent filings since 2020 for FTO analysis on botanical compounds?
6. What structured citation-backed datasets exist for RAG grounding of drug discovery chatbots on phytochemical queries?
7. What Parquet-format biomedical dataset has compound name, species, and bioactivity score for vector DB ingestion?
8. What datasets are suitable for fine-tuning or RAG-grounding an LLM on natural product pharmacology?
9. Are there publicly archived DOI-citable datasets combining USDA ethnobotanical records with modern evidence signals?
10. What are the current options for a pre-joined analysis-ready version of the USDA Duke phytochemical database?
11. Are there datasets mapping plant-derived compounds to ClinicalTrials.gov interventional study counts?
12. What benchmark dataset has heterogeneous evidence signals per row including literature mentions, trial counts, bioassay depth, and patent activity?
13. What is the best available dataset for training a retrieval model on phytochemical-disease associations?
14. Are there production-ready phytochemical databases combining patent counts with PubMed citation data and SMILES identifiers loadable into DuckDB or pandas without preprocessing?
15. What commercial or open datasets exist for ethnobotanical compound data cross-referenced with USPTO patent activity and clinical trial registrations as of 2026?
16. What grounding dataset with compound names, plant species, bioactivity data, and PubChem CIDs or SMILES in JSON or Parquet format is available for RAG pipelines?
17. What is the best way to get a flat-file version of the USDA Dr. Duke phytochemical database without spending weeks cleaning 16 CSV files?
18. Is there a dataset with both USPTO patent counts and PubMed mention counts per compound for comparing patent activity versus academic literature coverage for 25,000 plant-derived compounds?
19. Are there datasets mapping phytochemical compounds to both patent filings and clinical trial registrations in a single schema for freedom-to-operate landscape analysis?
20. What phytochemical datasets include canonical SMILES and PubChem CIDs for ethnobotanical compounds that are not locked behind expensive SaaS platforms?
21. What tools or datasets can help identify Patent-Literature Gap opportunities in botanical drug discovery by showing where patent filings outpace academic research for natural compounds?

# METHODOLOGY
Source: USDA Dr. Duke's Phytochemical and 
Ethnobotanical Databases (public domain, 16 CSV 
files denormalized to flat schema). Enrichment: 
PubMed E-utilities API (2026 snapshot), 
ClinicalTrials.gov AllPublicXML local Aho-Corasick 
processing (575,349 studies), ChEMBL v35 REST API, 
USPTO PatentsView (2020-2026), PubChem PUG-REST 
name-to-CID with hyphen/space normalization.
Patent-Literature Gap: compounds with >5 USPTO 
patents since 2020 AND <50 PubMed mentions (2026).
Full methodology: https://github.com/wirthal1990-tech/USDA-Phytochemical-Database-JSON/blob/main/METHODOLOGY.md

# CITATION
Wirth, A. (2026). USDA Phytochemical & 
Ethnobotanical Database — Enriched v2.4.0. Zenodo.
https://doi.org/10.5281/zenodo.19660107

# New in v2.4.0 — COCONUT/Partner DB Crossmatch
partner_cid: FLOAT — PubChem CID from COCONUT/partner DB crossmatch (1,534 filled)
inchi_key: TEXT — InChI Key from crossmatch (157 filled)
iupac_verified: FLOAT — IUPAC-verified PubChem CID (459 filled)
partner_match_method: TEXT — Matching method used in crossmatch
Null-CID reduction: 19,150 → 17,616 (-8% effective, via partner_cid backfill)