§06 · DATA LAYER

A bill of materials. Audit it.

Roughly forty of them, each named, each with its license and the date it was last checked. If a claim in an answer can't be traced to one of these, it doesn't get made.

Primary data sources, what each provides, its license, and the date the license was last verified.
SourceWhat it providesLicenseVerified
PubMed · NCBI E-utilities37M records. Publication types, MeSH headings, abstracts, registry links.3 requests/sec without a key. In live testing, 5 of 10 rapid unkeyed requests came back 429.US Government work
ClinicalTrials.gov API v2Enrollment, masking, allocation, status, and baseline demographics.Registry race and ethnicity reporting hit 91.4% after it became a requirement. Publications sit near 43%. For the trials that link, the registry wins.US Government work
Crossref RESTDOI resolution and update notices, including retractions.One of three independent retraction checks. Each of the three misses cases the others catch.Open
Retraction Watch databaseThe retraction corpus, downloadable in bulk.About 65MB. Optional local download for retraction checking with no API dependency at all.CC0
Europe PMCFull-text search, annotations, and better trial-registry linkage than PubMed alone.Open
OpenAlexJournal-level signals used for the trust term. Never impact factor.Impact factor is deliberately absent. It's proprietary, has no free API, and a journal mean over a skewed distribution predicts an individual paper poorly.Data CC0 · free API key since early 2026
CPICPharmacogenomic guidelines, allele definitions, diplotype-to-phenotype tables.Bundled as a snapshot. PharmGKB and ClinPGx content is never bundled: CC BY-SA plus an explicit no-sale clause.CC0
gnomAD v4Population allele frequencies, with sub-population detail.The terms page is a JavaScript app that returns nothing to a fetcher, so this is verified by reputation and not by primary source. It gets read in a browser before anything is bundled.Open access · terms review outstanding

§06.1Five endpoints. No auth on any of them.The worked example

§06.1 · PROVENANCE

Every endpoint the homepage example hit.

The panel on the front page is not an illustration. These are the URLs it was built from, the day they were called, and the license each one publishes. Re-run them.

  • CPIC API (PostgREST)CC0 1.0 Universal (public domain dedication)
    https://api.cpicpgx.org/v1/

    No auth, no API key. Endpoints used: /pair_view, /diplotype, /guideline, /recommendation_view. CPIC asks that its logo and the "CPIC®" mark are NOT reproduced in marketing material — citing CPIC in body text is fine.

  • CPIC CYP2C19 Frequency Table (xlsx)CC0 1.0 Universal
    https://files.cpicpgx.org/data/report/current/frequency/CYP2C19_frequency_table.xlsx

    HTTP 200, 70,464 bytes. Sheets: Allele frequency · Diplotype frequency · Phenotype frequency · References · Methods and Caveats · Notes · Change log. Latest change-log entry 2022-03-11 ("fixed diplotype/phenotype calculations").

  • gnomAD v4 GraphQL API
    https://gnomad.broadinstitute.org/api

    POST GraphQL, no auth, dataset id `gnomad_r4`. Rate limit 30 req/min/IP, max query cost 25 — batch with aliases and cache. gnomAD v4 is on GRCh38, so all coordinates below are GRCh38.

  • NCBI PubMed E-utilities
    https://eutils.ncbi.nlm.nih.gov/entrez/eutils/

    efetch db=pubmed for records; esearch db=pubmed for the screening counts.

  • Lee CR et al. CPIC Guideline for CYP2C19 Genotype and Clopidogrel Therapy: 2022 Update. Clin Pharmacol Ther 2022;112(5):959-967.
    https://pubmed.ncbi.nlm.nih.gov/35034351/

    PMID 35034351 · doi:10.1002/cpt.2526 · full text read via PMC9287492.


§06.2Named, not papered over.

§06.2 · THE HOLES

The holes, since you'd find them anyway

Four things a reader would look for and not find, each with the reason it is missing. A source list without this section is a marketing page.

  1. No free, commercially safe drug-interaction corpus exists. RxNorm's interaction API was permanently retired on 2024-01-02 with no replacement, and DDInter 2.0 is CC BY-NC-SA. So Biocite ships no interaction checker.
  2. Cochrane and Epistemonikos have no open API. Their systematic reviews reach the pipeline only through PubMed indexing.
  3. There's no authoritative machine-readable source for sex- or ancestry-stratified lab reference ranges. CALIPER is closest and it's browser-only. So the vault stores whatever range the reporting lab printed, and says which lab it came from.
  4. There's no free authoritative predatory-journal blacklist. Cabells is paid and Beall's list has been defunct since 2017. The rubric inverts to whitelists instead: MEDLINE indexing, DOAJ presence, OpenAlex core.

§06.3A licensing requirement, not decoration.

§06.3 · ATTRIBUTION

What the licenses require us to say.

NLM and CPIC attribution is a condition of using those datasets, not a courtesy. It ships in the footer of every page and it is repeated here in full.

Attributed on every page

NLM PubMed · Europe PMC · ClinicalTrials.gov · CPIC (CC0) · openFDA · DSLD · Crossref

  • CPIC content is CC0. CPIC asks that its logo and the CPIC® mark are not reproduced in marketing material, so neither appears anywhere on this site; CPIC is credited in body text instead.
  • PharmGKB and ClinPGx content is never bundled. It is CC BY-SA with an explicit no-sale clause, which is incompatible with shipping a snapshot inside an installable package.
  • gnomAD’s terms page is a JavaScript application and returns nothing to a fetcher. That row is verified by reputation only and it is read in a browser before anything is bundled. It is listed in the unverified log on the method page for the same reason.
§12 · GET IT

Point it at your own data.

One command. No account, no upload, no key required.

Node at least 20 · MIT · no telemetry
$ claude mcp add biocite -e BIOCITE_VAULT="$HOME/health-vault" -- npx -y biocite-mcp