Sources and acknowledgements
About › Sources & acknowledgements
Everything on this site is derived from public archives and open literature; this page names those sources and their terms, the software and models the pipeline uses, the funding, and the other projects that curate human-microbiome metadata — with an honest assessment of what each could add to the catalog and how it would join.
About this page
Last reviewed 2026-09-28.
Development of Microbiome Repo is supported by a grant from Anthropic to the Olm Lab (University of Colorado Boulder) for organising human microbiome metadata.
Data sources
| source | what we take from it | terms |
|---|---|---|
| European Nucleotide Archive (ENA) / INSDC (SRA, DDBJ) | Study, sample, run and BioSample records and attributes for every registry study (portal API); the archive metadata is route R1. | INSDC: free and unrestricted access to all records. |
| NCBI BioProject and BioSample | BioProject records (titles, descriptions, publications, organisations) and BioSample attributes; E-utilities. | NCBI public-domain data policy. |
| Europe PMC | Paper ↔ accession links, abstracts, open-access full text (JATS) and supplementary files behind routes R2, R3 and R4; author lists. | Europe PMC open-access subset; individual article licences apply to reuse of text. |
| Sandpiper (SingleM community profiles of public metagenomes) | Per-run community profiles and the count of profiled runs per study; profiles shown for the infant studies. | CC-BY (Woodcroft et al. 2025). |
| curatedMetagenomicData (cMD) | Gold standard for the infant-field extraction evaluation (age, delivery mode, preterm status precision) and the 22 infant BioProjects used for study-level recall. | Artistic-2.0 (Bioconductor package); data via ExperimentHub. |
| UBERON | Anchors of the body-site vocabulary (config/vocab/body_sites.yaml). | CC-BY 3.0. |
| GTDB | Taxonomy of the Sandpiper community panels (release stated on each panel). | CC-BY-SA 4.0. |
Software and models
| component | role | licence |
|---|---|---|
| DuckDB-WASM | In-browser SQL over the parquet tables (sample and registry explorers); pinned 1.29.0 from jsDelivr. | MIT. |
| DataTables and jQuery | Sortable study, cohort and scope tables. | MIT. |
| Apache Parquet / Arrow, pandas, DuckDB, Jinja2, Python-Markdown | Pipeline and site generator (microbiome_repo-pipeline). | Apache-2.0 / BSD / MIT (see the repository). |
| Anthropic Claude models | Model-assisted classification (registry), key triage (R1), value normalisation, cohort-statement extraction (R3, R4) and adjudication — always with a verbatim evidence quote; roles and model ids in config/models.yaml and on every model-derived row. | Anthropic usage policies. |
External curated resources incorporated
Each resource, what it offers, how we join it and the projects that use it: External resources.
Other groups have curated human microbiome metadata before us. Where their per-sample or per-study annotations join our samples through archive accessions, we incorporate them as route R2 evidence rows — evidence_source = external.<resource>, the locator a link to the record at the source, the quote the source's raw value — after the study's own supplementary tables in precedence (a sample's own archive attribute always wins). We are grateful to these projects; please cite them when you use values that carry their evidence source.
Acknowledgement. We gratefully acknowledge the teams behind African Human Microbiome Portal (H3ABioNet), BugSigDB, curatedMetagenomicData 3 (cMD 3), GMrepo v3, HMP / iHMP data portal, HumanMetagenomeDB, mBodyMap, Meta2DB (LLNL), MicrobeAtlas, MicrobiomeDB (VEuPathDB): their curation work is incorporated here as route-R2 evidence, every incorporated value links back to its source record (evidence_source = external.<resource>), and users of those values should cite the reference listed for the resource alongside this catalog. Resources that publish no licence are redistributed with attribution only; if you maintain one of them and prefer different terms, open an issue and we will comply.
| resource | what we incorporated | reference · terms |
|---|---|---|
| African Human Microbiome Portal (H3ABioNet) external.ahmp | 6,209 R2 evidence rows per-sample values for 2,192 samples in 15 catalog studies — country 2,192, health_condition 1,992, health_condition_detail 1,992, age_at_collection_days 33. Joined on archive accessions; each row links to the source record. | Database 2024 No formal licence; portal disclaimer: research purposes only. Export whole_data.tar.gz 2024-05-30. |
| BugSigDB external.bugsigdb | 103 R2 evidence rows study-level statements (scope study_all) for 0 samples in 62 catalog studies — country 55, antibiotic_exposure 36, health_condition 6, health_condition_detail 6. Joined on archive accessions; each row links to the source record. | Geistlinger et al. 2023 Nat Biotechnol CC-BY-4.0 (BugSigDBExports full dump 2026-09-28). |
| curatedMetagenomicData 3 (cMD 3) external.cmd | 141,320 R2 evidence rows per-sample values for 24,547 samples in 118 catalog studies — country 24,413, health_condition 23,576, health_condition_detail 23,089, sex 21,691, antibiotic_exposure 17,708, age_at_collection_days 16,757, bmi 14,086. Joined on archive accessions; each row links to the source record. | Pasolli et al. 2017 Nat Methods; cMD 3 — Nature Communications 2025 Artistic-2.0 (repository DESCRIPTION); source: curatedMetagenomicDataCuration commit 8059a76 (2026-09-28). |
| GMrepo v3 external.gmrepo | 140,605 R2 evidence rows per-sample values for 39,805 samples in 286 catalog studies — health_condition_detail 38,059, health_condition 37,165, country 35,361, sex 15,826, age_at_collection_days 11,600, bmi 2,521, antibiotic_exposure 73. Joined on archive accessions; each row links to the source record. Run-level phenotypes from the live v3 API (MeSH terms mapped to the health-condition vocabulary). | Wu et al. 2020, Dai et al. 2022, Liu et al. 2026 NAR (v3) No licence published; the site asks for citation of Dai et al. 2022 (GMrepo v2). Live API, 2026-09-28. |
| HMP / iHMP data portal external.hmp_ihmp | 14,152 R2 evidence rows per-sample values for 2,456 samples in 3 catalog studies — health_condition 2,456, health_condition_detail 2,456, sex 2,317, antibiotic_exposure 2,317, country 2,317, age_at_collection_days 2,289. Joined on archive accessions; each row links to the source record. The HMP data portal API is retired; values come from the IBDMDB public metadata file and the archived HMASM-690 table. | Human Microbiome Project Consortium 2012; iHMP 2019 IBDMDB hmp2_metadata_2018-08-20.csv (public) and the archived HMP DACC HMASM-690 table (Wayback 2024-04-15); portal API retired. |
| HumanMetagenomeDB external.hmgdb | 34,330 R2 evidence rows per-sample values for 19,876 samples in 165 catalog studies — country 17,135, sex 5,942, age_at_collection_days 4,485, health_condition 2,709, health_condition_detail 2,708, bmi 1,199, antibiotic_exposure 152. Joined on archive accessions; each row links to the source record. | Kasmanas et al. 2021 NAR No licence statement published (© HumanMetagenomeDB & TerrestrialMetagenomeDB team). Release 1.1 (June 2020), fetched 2026-09-28. |
| mBodyMap external.mbodymap | 6,404 R2 evidence rows per-sample values for 3,070 samples in 5 catalog studies — health_condition 3,070, sex 3,062, health_condition_detail 272. Joined on archive accessions; each row links to the source record. | Jin et al. 2022 NAR No licence statement; the Help page asks for citation. Download files dated 2021-09-26. |
| Meta2DB (LLNL) external.meta2db | 51,784 R2 evidence rows per-sample values for 11,338 samples in 78 catalog studies — health_condition 10,559, country 9,809, sex 7,969, age_at_collection_days 7,059, antibiotic_exposure 5,776, bmi 5,406, health_condition_detail 5,206. Joined on archive accessions; each row links to the source record. | Meta2DB 2024 bioRxiv / 2026 CC-BY-4.0 (Zenodo record 17315984, 2025-10-13). |
| MicrobeAtlas external.microbeatlas | 15,074 R2 evidence rows per-sample values for 15,074 samples in 151 catalog studies — country 15,074. Joined on archive accessions; each row links to the source record. Country derived from the source's parsed coordinates (reverse geocoding, confidence 0.65); 17 studies with systematic disagreement against archive attributes were excluded. | Cell 2026 No licence statement found on the site; citation requested. Bulk metadata 2026-01-29. |
| MicrobiomeDB (VEuPathDB) external.microbiomedb | 18,921 R2 evidence rows per-sample values for 5,113 samples in 10 catalog studies — age_at_collection_days 5,021, sex 4,501, country 3,405, antibiotic_exposure 2,528, health_condition 1,577, health_condition_detail 1,455, bmi 434. Joined on archive accessions; each row links to the source record. | Oliveira et al. 2018 NAR VEuPathDB data-use statement: data provided freely for public use, cite the resource. Build 37 (May 2024). |
Also consulted, not incorporated (out of scope or no per-sample archive-linked metadata): EMBERS (LLM extraction of metadata from 26,435 gut-microbiome papers) · Human Microbiome Compendium (MicroBioMap) · MGnify (EBI Metagenomics) · PRIME · gutMEGA, gutMDisorder, Disbiome, HMDAD, GutMetaNet, HGMT.
Citing and licence
Concept DOI (all versions, resolves to the latest archived release): 10.5281/zenodo.23087712.
Curated tables (registry_*, gut_* and the infant extension) are released under CC-BY-4.0; archive metadata remains under INSDC terms; community profiles from Sandpiper (CC-BY). When you use values from this catalog, please also cite the primary studies — every study page lists its papers.