Open data behind OnCo
18 sources already feed the corpus (10 by script or live query, the rest read by hand and cited); 79 more are mapped as candidates across 12 themes. Every row says what it holds, under which licence, and what it would fuel.
How data gets in
- A script under
scripts/queries a public API with a namedUser-Agentand a contact address, one request at a time with backoff, and writes a compact snapshot underpublic/. Each snapshot carriessource,fetchedand, where the source states one,license. - GitHub Actions rerun the trial, literature and fact-check scripts weekly and open a pull request with the diff, so every refresh is reviewed before it lands. Other snapshots are refreshed by hand with
npm run fetch:*. - The build reads the snapshots and the hand-written corpus in
src/data/, validates every id and reference, and renders static pages and the JSON API. Nothing is fetched at request time; a few components query ClinicalTrials.gov, Europe PMC or Nominatim live in the browser when a reader asks. npm run factcheckcross-checks recorded approvals and trial statuses against openFDA and ClinicalTrials.gov; mismatches appear on the Audit page.
Attribution rules
- We store counts, identifiers and short derived records, never wholesale copies of a database. Every page links back to the primary record.
- Sources are named on the page that uses them and in the snapshot file; molecule files keep PubChem and RCSB terms, logos remain trademarks and carry per-file Wikimedia Commons licences.
- Sources with academic-only or commercial licences (COSMIC, OncoKB, DrugBank, KEGG, NCCN text) are linked, not ingested, unless their terms allow redistribution under CC BY 4.0.
- No patient-level data, no personal data beyond public professional profiles, and no scraping of patient forums or social media. Epidemiology is aggregate estimates only.
- The corpus is CC BY 4.0: name OnCo and link to onco.cc; upstream licences stay attached to what came from upstream. Details in
docs/DATA-SOURCES.md.
In use
18 sourcesWhat we already pull from, what we take, the script or component that reads it, and how often. Click a chip to filter; the name opens the OnCo collection page when there is one.
18 sources
| Fuels | How it gets in | Access | Licence | Site | ||
|---|---|---|---|---|---|---|
cBioPortal (linked for prevalence) Genomics and targets Alteration frequencies per cancer type for the /prevalence/ matrix, read from study summaries. The REST API could compute them per gene and study. | Biomarker prevalence rows on targets cite cBioPortal study views (47 links); no script yet | By hand | docs.cbioportal.org | |||
Clarivate Journal Citation Reports (impact factors) Literature One number per journal where it is common knowledge; omitted otherwise. | src/data/journals.ts impactFactor field, sourced to the 2024 JCR release | Annual, by hand | clarivate.com | |||
ClinicalTrials.gov API v2 Trials and results We take NCT id, title, status, phases, conditions, start and primary-completion dates and lead sponsor for phase 2/3 studies matching each product name, plus status checks for our own trial records and study counts per country. Results sections, arms, eligibility and outcome measures are not pulled yet. | scripts/fetch-trials.ts (phase 2/3 studies per product → public/trials/), scripts/factcheck.ts (overallStatus per NCT), scripts/fetch-countries.ts (studies per location country); TrialFinder and TrialFinderGeo query it live in the browser | Weekly (refresh-trials.yml, Mondays; factcheck.yml, Mondays); countries by hand | clinicaltrials.gov | |||
DailyMed (NLM prescribing information) Regulatory Toxicity tables and dosing on product pages come from US labels; the field is omitted rather than estimated when a label gives no number. A REST and SPL-XML service exists and could replace the manual reading. | src/data/drug-depth.ts: grade 3+ adverse-event rates and dosing read from the label, each row linking the DailyMed setid | By hand when a label changes | dailymed.nlm.nih.gov | |||
EMA European public assessment reports Regulatory One region matrix for approved and late-stage products across six regulators. EMA publishes a nightly Excel table of all authorised medicines and twice-daily JSON dumps of the website that could automate the EU column. | src/data/regional-approvals.ts: EU rows checked against EPAR pages are marked verified; UK, Japan, China and Australia rows link regulator searches (MHRA products, PMDA, NMPA, TGA ARTG) | By hand; EU rows verified 2026-09-08 | ema.europa.eu | |||
Europe PMC REST API Literature Hit counts per year 2019-2026, last-12 versus prior-12 months and the five most recent titles (title, DOI, PMID, journal, date, citations) for every drug, target, cancer and technology with a usable query; /papers/ ranks fastest-growing topics from the snapshot. We store counts and citation-level metadata only, never abstracts or full text. | scripts/fetch-papers.ts → public/papers/<id>.json and index.json; LatestPapers queries it live on entity pages | Weekly (refresh-papers.yml, Tuesdays, --force) | europepmc.org | |||
FDA Oncology Center of Excellence approvals and Drugs@FDA Regulatory The official record of what is approved for what in the US. The Drugs@FDA data files and openFDA drugsfda endpoint would let a script check every date we hold. | Approval dates, indications and CRLs are written into drugs.ts, drug-depth.ts and regional-approvals.ts from the notification pages and accessdata.fda.gov letters (108 fda.gov and 24 accessdata links in the corpus) | By hand, continuously | fda.gov | |||
GLOBOCAN 2022 (IARC Global Cancer Observatory) Epidemiology New cases, incidence ASR, deaths, mortality ASR and cumulative risk to 74 for 185 countries and 36 cancer sites, both sexes, all ages, plus the world aggregate; every page states these are estimates, not registry counts, and prints the IARC citation. Cancers without a GCO site (or sharing a site) are disclosed on /cases/. | scripts/fetch-globocan.ts (gco-api.iarc.fr, the endpoint the GCO site itself uses) → public/globocan/countries.json; src/lib/globocan.ts and src/data/globocan-map.ts map OnCo cancers to GCO site codes | By hand (npm run fetch:globocan); one edition, 2022 | gco.iarc.who.int | |||
NCCN and ESMO guidelines (linked, not ingested) Care and quality We record the category and link the guideline; guideline text is never copied. | Every standard-of-care row in the cancer records carries a guideline mapping (NCCN category, ESMO-MCBS grade) with a link; 137 nccn.org links in the corpus | By hand with each guideline version | nccn.org | |||
Newsweek/Statista, Nature Index and SCImago rankings Institutions and people Inputs to the institution table, each with its methodology linked; the formula is printed under the table. | Institution records carry newsweekOncology2026 and NCI designation; src/lib/ranking.ts combines them with graph links into a disclosed score | Annual, by hand | rankings.newsweek.com | |||
OpenAlex Institutions and people Institution ids resolved by search with manual overrides, then work counts filtered to primary topic subfield Oncology with the institution lineage filter; country output uses whole counting so collaborations credit every country. Feeds /universities/ and /countries/. | scripts/fetch-openalex.ts (oncology works and citations per institution, subfield 2730, 2024-2025) → public/openalex/institutions.json; scripts/fetch-countries.ts (works per author country 2021-2025, highly cited, open access) → public/openalex/countries.json | By hand (npm run fetch:openalex, fetch:countries) | help.openalex.org | |||
openFDA drug label endpoint Regulatory Existence of a US prescribing-information record per product; 130 products checked last run. The drugsfda, NDC and adverse-event endpoints are not used yet. | scripts/factcheck.ts searches openfda.generic_name for every US-approved product and flags products with no label (or a label but no recorded approval) → public/factcheck.json, shown on /audit/ | Weekly (factcheck.yml, Mondays) | open.fda.gov | |||
OpenStreetMap Nominatim Geography Nothing stored; the browser calls Nominatim once when a reader types a place, and the result stays on the device. | src/lib/ctgov-geo.ts geocodes a reader's town on explicit action only, then asks ClinicalTrials.gov for recruiting studies within a radius | Live, on user action | operations.osmfoundation.org | |||
PubChem PUG REST Structures and chemistry Every small-molecule product, payload and linker with a structure definition gets a self-hosted rotating wireframe. Biologics, cells and devices show an explained placeholder instead. | scripts/fetch-structures.ts resolves src/data/structures.ts by name or CID, downloads the 3D (or 2D) SDF and writes compact atom and bond arrays to public/structures/pubchem-<cid>.json | By hand (npm run fetch:structures); idempotent | pubchem.ncbi.nlm.nih.gov | |||
RCSB Protein Data Bank Structures and chemistry Antibody-antigen complexes and drug-target co-crystals for products without a small-molecule structure, for example sacituzumab Fab bound to TROP2. First model only, hydrogens dropped, solvent and buffer residues skipped. No request-rate limit is set by RCSB. | scripts/fetch-structures.ts downloads files.rcsb.org/download/<id>.pdb and keeps the C-alpha trace, drug-like ligands and pocket residues → public/structures/pdb-<id>.json | By hand (npm run fetch:structures) | rcsb.org | |||
Wikidata and Wikimedia Commons Community and reference 441 self-hosted logo files for companies, institutions and collections; Clearbit and Google favicons are fallbacks, not open data, and are labelled as such in the index. Wikipedia links sit on most glossary terms and many entities. | scripts/fetch-logos.ts searches Wikidata by name or website (P856 domain match, P31 organisation class), takes the P154 logo from Commons with its extmetadata licence → public/logos/; entity `wikipedia` fields link the encyclopaedia | By hand (npm run fetch:logos); idempotent | wikidata.org | |||
World Bank population indicators Epidemiology Denominators for per-capita research output on /countries/. The World Bank Indicators API could refresh this in one call. | src/data/country-extras.ts: 2024 population estimates per country, rounded | By hand | data.worldbank.org | |||
world-atlas (Natural Earth) Geography Country outlines for the institutions map and the GLOBOCAN choropleth. | src/components/WorldMap.tsx loads countries-110m.json from jsDelivr at runtime and joins on ISO numeric codes | Pinned package version | github.com |
Candidates
79 sources · 39 fully open · 20 with registration, academic or commercial termsWhat we could extend to, grouped by what it would fuel. Effort is a first estimate for a working fetch script and the fields it would add; risks name the licence, rate-limit or personal-data issues to settle first.
79 candidate sources
| Would fuel | Access | Licence | Risks | Site | |||
|---|---|---|---|---|---|---|---|
CMS Care Compare provider data Care and quality Hospital-level quality and patient-experience measures with CMS certification numbers via a DKAN REST and SQL API, no key; a few oncology-relevant measures (for example, chemotherapy in the last two weeks of life for PPS-exempt cancer hospitals). | Few cancer-specific measures; provider ids to institution mapping. | data.cms.gov | |||||
COSD (Cancer Outcomes and Services Dataset, England) Care and quality The national data standard behind NDRS outputs; OnCo would use only the published aggregate tables (see the ONS and NDRS row). | Patient-level data are out of scope for a public corpus. | digital.nhs.uk | |||||
Getting It Right First Time (GIRFT) Care and quality Specialty reports on unwarranted variation, including urology, breast surgery, radiotherapy and pathology. | PDF narrative reports; cite, do not ingest. | gettingitrightfirsttime.co.uk | |||||
NCDB National Cancer Database Care and quality Seventy percent of US cancer diagnoses with treatment and outcome; only aggregate public reports could be cited. | Patient-level data closed; hospital identifiers confidential; AI-use prohibition in site terms. | facs.org | |||||
NHS England cancer waiting times Care and quality Monthly provider-level performance against the Faster Diagnosis, 31-day and 62-day standards as Excel and CSV; a live indicator for the diagnostic-delay bottleneck. | Spreadsheet layout changes; ICB boundary changes; England only. | england.nhs.uk | |||||
UK national cancer audits (NATCAN) Care and quality Ten national audits (lung, bowel, prostate, breast, ovarian, pancreatic, kidney, oesophago-gastric, non-Hodgkin lymphoma, metastatic breast) with trust-level indicators in reports and dashboards. | Reuse needs permission; PDF reports; trust names need mapping to institutions. | natcan.org.uk | |||||
Patient forums and social media (Reddit, Facebook groups, X) Community and reference Listed to record the decision: OnCo does not ingest, quote or analyse patient forum or social media posts. They contain health information about identifiable people who did not consent to reuse; the site links to patient organisations instead. | Personal health data, consent, platform terms; excluded by policy. | reddit.com | |||||
Wikidata SPARQL (companies, drugs, people) Community and reference Structured facts with external identifiers (ChEMBL, DrugBank, PubChem, ROR, GRID, ORCID, ISNI) for one-query enrichment and identifier crosswalks. | Endpoint limits: 60-second queries, five parallel per IP, mandatory User-Agent; people data must stay to public professional facts. | query.wikidata.org | |||||
Wikipedia REST API (summaries) Community and reference Page summaries and lead images for the Wikipedia links we already hold; the REST summary endpoint is fast and cached. Wikimedia's User-Agent and API etiquette policies apply. | Share-alike text: display as attributed quotation, not as OnCo's own TL;DR. | en.wikipedia.org | |||||
ClinicalTrials.gov sponsor and collaborator fields Companies and pipelines Aggregating the sponsorCollaboratorsModule and location facilities we already receive gives each company an active-trial count by phase and each institution a trial-site count. | Sponsor naming variants (already handled by an alias table for our trial rows). | clinicaltrials.gov | |||||
EPO Open Patent Services (OPS) Companies and pipelines Bibliographic and legal-status data for European and worldwide patents, including supplementary protection certificates that set exclusivity dates in Europe. | OAuth token flow; throttling; XML responses. | epo.org | |||||
Google Patents Public Datasets on BigQuery Companies and pipelines Worldwide patent publications with families, assignees and CPC codes; trend charts for ADC, radioligand and CAR-T patenting by year and country. First terabyte of queries a month is free. | Needs a Google Cloud project and query costs; repository archived; heavy for a static site pipeline. | github.com | |||||
OpenCorporates Companies and pipelines Cross-jurisdiction company records and parent-subsidiary links (useful for acquisitions like Seagen into Pfizer). | Share-alike obligations; API key and quotas; scraping prohibited. | opencorporates.com | |||||
SEC EDGAR APIs and full-text search Companies and pipelines Company facts (XBRL) give revenue and R&D per year; 10-K pipeline tables and 8-K press releases give dated approvals, readouts and partnerships for listed sponsors. | Free-text extraction from filings; US-listed companies only; undeclared bots are blocked. | sec.gov | |||||
UK Companies House API Companies and pipelines Company numbers, status and filed accounts for UK biotechs and charities (Cancer Research UK is a registered company). | Officer and PSC data are personal data; use company-level facts only. | developer.company-information.service.gov.uk | |||||
USPTO PatentsView (moving to the Open Data Portal) Companies and pipelines Granted US patents with assignees, inventors, CPC classes and citations. The PatentSearch API is migrating to the USPTO Open Data Portal: the old search.patentsview.org host no longer resolves and new keys need a USPTO.gov account with multi-factor authentication. | Service in flux; inventor names are personal data (public record); linking patents to products needs the Orange Book or manual curation. | data.uspto.gov | |||||
CDC WONDER Epidemiology Mortality and United States Cancer Statistics queries as XML over HTTP; the API returns national-level vital statistics only, sub-national groupings are web-tool only. | No state-level data via the API; guidance of one query every two minutes; suppression of counts under ten. | wonder.cdc.gov | |||||
CONCORD programme (global survival surveillance) Epidemiology Five-year net survival for 18 cancers in 71 countries, the standard for cross-country survival comparison; figures live in supplementary tables. | Data by request only; would be transcription of published tables with citation. | csg.lshtm.ac.uk | |||||
ECIS European Cancer Information System Epidemiology EU-27 estimates and registry-based indicators with data exports; cross-check for GLOBOCAN European figures. | Exports from an interactive tool behind a web application firewall; confirm the ECIS-specific notice by hand. | ecis.jrc.ec.europa.eu | |||||
GCO Cancer Tomorrow, Cancer Over Time and Cancer Survival Epidemiology Same API family as the Cancer Today data we hold: projected cases to 2050 by country and cancer, registry-based time trends, and SURVCAN survival for lower-income registries. | Undocumented endpoints that refuse anonymous clients; projections rest on demographic assumptions and must be labelled. | gco.iarc.who.int | |||||
NCI SEER Epidemiology SEER*Explorer publishes incidence and survival by site, stage, sex, race and year; the SEER API (api.seer.cancer.gov) exposes reference datasets rather than case data. Would give every cancer page a sourced stage-at-diagnosis and survival panel, most simply from the published tables. | US only; case-level data is off limits for a public corpus; acknowledgement of the SEER database version required. | seer.cancer.gov | |||||
NORDCAN Epidemiology Registry data for the Nordic countries with survival trends since the 1960s; a survival benchmark for the annual report. | No documented API; the site is an interactive app; non-commercial clause. | nordcan.iarc.fr | |||||
ONS and NHS England NDRS cancer statistics Epidemiology Registry-grade incidence, survival by stage, routes to diagnosis and Cancer Outcomes and Services Dataset outputs as spreadsheets; the best public national registry data anywhere. | Bot protection means manual downloads; spreadsheets with changing layouts; England only. | digital.nhs.uk | |||||
cBioPortal web API Genomics and targets Mutation, copy-number and fusion frequencies per gene across curated study sets, computed rather than transcribed, with sample counts to show uncertainty. OpenAPI-documented REST, no key for the public instance. | Per-study terms (MSK-IMPACT and GENIE citations); heavy queries; no rate limit published. | docs.cbioportal.org | |||||
ChEMBL Genomics and targets Drug mechanisms with action type and target, approval year, ATC codes, indications by phase and molecule properties for every compound; also holds ADC payload chemistry. Share-alike is compatible if the derived table is published under a compatible licence. | CC BY-SA obliges share-alike on derived data tables; ADC records are patchy. | chembl.gitbook.io | |||||
CIViC Genomics and targets Community-curated evidence items linking variant, disease, drug and evidence level to a PubMed id, fully open. The natural open substitute for OncoKB in the tumour board: evidence counts per biomarker and drug with citations. GraphQL at civicdb.org/api/graphql, 3 requests a second without a key, unlimited with a free one. | Curation coverage is uneven; evidence levels differ from OncoKB's. | civicdb.org | |||||
ClinVar Genomics and targets Pathogenicity classifications and review status per germline variant via E-utilities or weekly XML and monthly VCF; would back the hereditary-syndrome terms and germline biomarkers with counts of pathogenic variants per gene. | Variant-level detail is beyond OnCo's scope; keep to gene-level summaries. | ncbi.nlm.nih.gov | |||||
COSMIC Genomics and targets Cancer Gene Census, mutation hotspots and signatures. OnCo links to COSMIC pages; ingesting data into a public website is explicitly a licensed use. | Licence bars redistribution and public-website use; link only. | cosmickb.org | |||||
DepMap Genomics and targets Quarterly CRISPR gene-effect and drug-sensitivity matrices; a per-target summary (fraction of lines dependent, lineage selectivity) would give every target page a druggability signal. | Licence to confirm on the release page; large files (gigabytes) need offline aggregation, not a build-time fetch. | depmap.org | |||||
DrugBank Genomics and targets Rich pharmacology and drug-interaction data. The non-commercial clause conflicts with CC BY 4.0 redistribution, so OnCo would use only the CC0 vocabulary (names and identifiers) and link the rest. | Licence bars redistribution of the main dataset; link only. | go.drugbank.com | |||||
GTEx Portal Genomics and targets Median TPM per gene across 54 normal tissues via the v2 REST API (no key); paired with HPA it explains ADC toxicity patterns on target pages. | Redistributors must keep data current or flag staleness; bulk RNA in adults, not protein. | gtexportal.org | |||||
Human Protein Atlas Genomics and targets Per-gene JSON, XML and TSV (proteinatlas.org/<ENSG>.json) with tissue specificity, cancer staining summaries and prognostic associations; the key data for whether an ADC target is safe. | IHC is semi-quantitative; cite the version. | proteinatlas.org | |||||
KEGG Genomics and targets Classic pathway maps and disease entries. Licence forbids non-academic use and redistribution, so OnCo links KEGG pathway ids rather than storing content. | Licence bars redistribution; link only. | kegg.jp | |||||
NCI Genomic Data Commons API Genomics and targets Simple somatic mutation frequencies and clinical variables per project (TCGA-BRCA and so on) via the filter API; the reference for TCGA-based prevalence numbers. | Open tier only; large responses; harmonisation versions change numbers. | gdc.cancer.gov | |||||
OncoKB Genomics and targets FDA-recognised levels of evidence per alteration and drug. The public actionable-genes page can be cited with links; API data cannot be republished without a licence. | Licence bars redistribution; link only or negotiate terms. | oncokb.org | |||||
Open Targets Platform Genomics and targets One GraphQL call per target returns association scores by disease with evidence counts, tractability buckets, known drugs with phases and safety liabilities. The highest-yield single addition for the targets kind. | Open Targets discourages entity-by-entity querying and points bulk users to downloads or BigQuery; EFO disease ids need mapping to OnCo cancers; scores change between releases. | platform-docs.opentargets.org | |||||
Reactome Genomics and targets Curated human pathways with participants and diagrams via the ContentService; would let /pathways/ link each hand-drawn diagram to the canonical reaction list and pull member genes. | Granularity is finer than OnCo's diagrams; mapping by hand. | reactome.org | |||||
STRING Genomics and targets Protein interaction partners with confidence scores per gene; a short 'interacts with' list on target pages that cross-links to other OnCo targets. API asks for a caller_identity and one second between calls; pin the version URL. | API is for occasional access, bulk via downloads; interaction lists are long, cap and threshold. | string-db.org | |||||
UniProt Genomics and targets Curated function text, subcellular location and cross-references per protein via rest.uniprot.org; a UniProt accession on each target would anchor every other genomics join. | Some entries may be covered by patents or other rights; otherwise none significant. | uniprot.org | |||||
NCI Imaging Data Commons (IDC) Imaging and pathology Cloud-hosted radiology and pathology collections with BigQuery metadata tables (licence per series) and REST endpoints for licences and citations; counts of cases per cancer and modality for a dataset finder. | BigQuery costs; mixed-licence collections; de-identified patient data, index only. | learn.canceridc.dev | |||||
The Cancer Imaging Archive (TCIA) Imaging and pathology NBIA REST API (no key) lists collections, modalities and series counts; OnCo would index collections per cancer and modality with their licences, not images. | De-identified patient data: index metadata only, never host images; no re-identification or facial reconstruction. | cancerimagingarchive.net | |||||
EU CORDIS (Horizon Europe and Horizon 2020 projects) Institutions and people Bulk CSV, XLSX and JSON of projects, participants and funding under the health cluster and the Cancer Mission. | Organisation names need mapping to institutions; contact names in beneficiary records. | data.europa.eu | |||||
GRID (Global Research Identifier Database, retired) Institutions and people Retired in 2021 and folded into ROR; only useful for resolving legacy GRID ids in older datasets. | No updates; use ROR instead. | grid.ac | |||||
NIH RePORTER API Institutions and people Projects, funding amounts, organisations and PIs with NCI as funding institute, searchable by text and organisation, no key; sums per institution and per cancer would make /funding/ computed rather than transcribed. | One request a second, large jobs at night US time; PI names are personal data (public record). | api.reporter.nih.gov | |||||
OpenAlex institutions and funders Institutions and people We resolve ids and count works; the institution entity also carries ROR, geo, type, homepage and image URLs, and the funders entity would link grant funders to output. | Daily usage budget; otherwise none. | help.openalex.org | |||||
ORCID public API Institutions and people Affiliations, works and identifiers for researchers who keep public records; the schema already has an orcid field on people. Registered clients get 12 requests a second and 100,000 reads a day. | Personal data: use only what the researcher made public, link rather than copy, no unsolicited contact. | info.orcid.org | |||||
Research Organization Registry (ROR) Institutions and people A ROR id per institution would make joins to OpenAlex, Crossref funders and ORCID exact instead of name-based. REST v2 with no key, 2,000 requests per five minutes; full dump on Zenodo. | None significant. | ror.readme.io | |||||
UKRI Gateway to Research Institutions and people Projects, funders, organisations and outcomes for UK public research funding as XML or JSON, no key. | Named principal investigators; CRUK's own grants are not included. | gtr.ukri.org | |||||
bioRxiv and medRxiv API Literature Preprints by date range and category with published-DOI links, 100 records a call; Europe PMC already indexes them (SRC:PPR), so this is mainly for a faster daily pulse. | Unreviewed; must be labelled as preprints; full text reuse is per article. | api.biorxiv.org | |||||
Crossref REST API Literature Resolve every DOI in the corpus (622 links) to title, authors, journal, date, licence URL, funder ids and reference lists; funder ids connect papers to /funding/. Polite pool with a mailto parameter; 50 requests a second per the rate-limit headers. | Back off on HTTP 429; abstracts present only when publishers deposit them. | crossref.org | |||||
OpenAlex works, authors and topics Literature We use institutions and countries; works and authors would give citation counts per key paper, ORCID-linked author profiles for people, and topic trends finer than subfield 2730. | Daily usage budget (free tier about US$1 a day, ten times that with a free key); author disambiguation errors. | help.openalex.org | |||||
PubMed E-utilities Literature MeSH-tagged records and publication types (randomised controlled trial, meta-analysis) give a quality filter Europe PMC lacks; esearch counts by MeSH would sharpen the /papers/ growth measure. | 3 requests a second without a key, 10 with one; large jobs at night or weekends US time; XML responses. | ncbi.nlm.nih.gov | |||||
Semantic Scholar Academic Graph API Literature Citation contexts, influential-citation counts and machine TL;DRs per paper. The licence is not an open one, so any derived table would be display-only with the mandated attribution. | Non-commercial licence conflicts with CC BY 4.0 redistribution; unauthenticated pool is shared; TL;DRs are machine-generated and must be labelled. | semanticscholar.org | |||||
Unpaywall API Literature Open-access status and best free full-text location per DOI, so every key paper can show a legal free link when one exists. | Licence to confirm before storing results; commercial data feed is paid. | unpaywall.org | |||||
Australian PBS Public Data API Pricing and access REST JSON v3 with the monthly schedule (items, restrictions, dispensed prices); the legacy XML files were discontinued. PBAC outcomes give the reimbursement decision behind each listing. | Throttle of one request per 20 seconds shared by all users makes bulk pulls impractical; licence to confirm. | data.pbs.gov.au | |||||
CMS Medicare Part B ASP drug pricing files Pricing and access Quarterly average sales price per HCPCS code for infused and injected drugs, where most oncology spend is, with an NDC-HCPCS crosswalk; history to 2005. Gives a sourced US price per dose for antibodies, ADCs and radioligands. | HCPCS to product mapping by hand; ASP plus 6% is the payment, not the list price. | cms.gov | |||||
IQWiG and G-BA benefit assessments (Germany) Pricing and access AMNOG added-benefit verdicts (major, considerable, minor, non-quantifiable, none) per indication, the most rigorous public comparative-effectiveness read on new cancer drugs. G-BA publishes bulk XML of decisions with an XSD on the 1st and 15th of each month and an RSS feed; IQWiG dossier assessments have English summaries. | Terms to read before redistribution; German-first documents; mapping subgroups to our indications. | g-ba.de | |||||
Medicaid NADAC (National Average Drug Acquisition Cost) Pricing and access Weekly pharmacy acquisition cost per NDC for retail drugs, the best public proxy for oral oncology drug prices; DKAN-style REST API and CSV downloads, no key. | Retail pharmacy only; hospital and specialty channels are missing. | data.medicaid.gov | |||||
Medicare Part D and Part B spending by drug Pricing and access Annual total spending, claims, beneficiaries and average spend per beneficiary by brand and generic name with manufacturer, via a JSON:API endpoint (pages of up to 5,000 rows, no key). One join on generic name gives every product page a US spend line and the annual report a top-spend table. | Two-year lag; small cells suppressed; brand-generic aggregation. | data.cms.gov | |||||
NICE Syndication API Pricing and access Every technology appraisal with recommendation status (recommended, optimised, Cancer Drugs Fund, not recommended), dates and indication text, plus guidance in development for the calendar. Turns UK access into a machine-checked column. | Registration and licence terms to confirm by hand; recommendation text is licensed for display without alteration. | nice.org.uk | |||||
Scottish Medicines Consortium advice Pricing and access Accepted, restricted or not-recommended decisions with dates and a monthly decision list, often earlier than NICE for the same drug. | Strict non-commercial clause; HTML and PDF only; cite decisions, do not copy text. | scottishmedicines.org.uk | |||||
Drugs@FDA data files Regulatory Zip of twelve tab-delimited tables (Applications, Products, Submissions, MarketingStatus and so on) updated each weekday morning. One download replaces thousands of API calls for a full reconciliation of application numbers and dates. | Schema documented only in a readme; brand and generic name matching. | fda.gov | |||||
EMA medicines data download and website JSON Regulatory Nightly Excel tables (medicines, procedures, referrals, orphan designations, shortages) and twice-daily JSON dumps of the whole website, including every medicine's authorisation status and date, therapeutic area, ATC code and EPAR URL. Would automate the EU column of the region matrix and put CHMP opinions on the calendar. | No query API; company-authored product information is outside the reuse permission; indications are free text. | ema.europa.eu | |||||
FDA Purple Book (biologics and biosimilars) Regulatory Monthly CSV and XLSX (2020 to present) of every licensed biologic with BLA number, licensure date, reference product, biosimilar and interchangeable status and exclusivity expiry. Fills the biologics gap that Drugs@FDA leaves and enables a biosimilar layer on antibody pages. | Products can be added to history retroactively; no API. | purplebooksearch.fda.gov | |||||
Health Canada Drug Product Database and Notice of Compliance Regulatory REST JSON with twelve endpoints and no key for every marketed drug (DIN, ingredient, status, company), plus text extracts of the Notice of Compliance database with approval dates and Project Orbis decisions. | Canada is not yet a column in the region matrix; adding it is a schema change. | health-products.canada.ca | |||||
MHRA Products and marketing authorisation data Regulatory UK product licences with SmPCs and public assessment reports; searchable, no documented API or bulk feed. | Undocumented site endpoints; UK dates since Brexit differ from EU; SmPC text is third-party. | products.mhra.gov.uk | |||||
NMPA and CDE (China) Regulatory Approval announcements, priority review and breakthrough designations for the Chinese pipeline, now the largest source of new oncology candidates. English NMPA databases are limited to vaccines and export certificates; drug detail is Chinese-only. | Chinese-language only, anti-bot protection, no API, data-export considerations; secondary trackers are commercial. | cde.org.cn | |||||
openFDA drugsfda and NDC endpoints Regulatory Application number, sponsor, product strengths, submission history with dates and types (approval, labelling supplement, efficacy supplement) per product. Would make every US approval date we hold machine-checked and add label-change events to the regulatory timeline automatically. | Same rate limits as the label endpoint; indication text lives in the label, not in drugsfda. | open.fda.gov | |||||
PMDA approved products list Regulatory One English PDF listing approved products since 2004 plus review reports, the richest public source on Japanese pivotal data. Because automated download is prohibited, the route is a periodic manual read of the PDF. | Automated crawling forbidden; English pages machine-translated; fiscal-year structure. | pmda.go.jp | |||||
TGA Australian Register of Therapeutic Goods Regulatory Searchable register with a public data extract on data.gov.au; AusPAR reports give approval dates and indications. | Licence to confirm by hand; the extract covers all therapeutic goods and needs ATC filtering. | tga.gov.au | |||||
AlphaFold Protein Structure Database Structures and chemistry Predicted structures by UniProt accession with per-residue confidence via a small REST API (no key); would give every target page a rotating wireframe where no experimental structure exists. | Predictions must be labelled as such and coloured by confidence; not validated for clinical use. | alphafold.ebi.ac.uk | |||||
PubChem compound properties and synonyms Structures and chemistry We fetch structures only; properties, synonyms and cross-references (ChEMBL, DrugBank, CAS) per CID would give each molecule page its identifiers and a molecular-weight line. | Same 5 requests a second policy. | pubchem.ncbi.nlm.nih.gov | |||||
ChiCTR (Chinese Clinical Trial Registry) Trials and results The registry behind the fast-growing Chinese ADC, bispecific and PD-1 pipeline; complements CDE. Also indexed by WHO ICTRP. | No API, anti-bot measures, Chinese-language records; scraping would breach site terms. | chictr.org.cn | |||||
ClinicalTrials.gov results sections and full study records Trials and results The same API we already call, asking for resultsSection (participant flow, baseline, outcome measures, adverse events), armsInterventionsModule, eligibilityModule and the dates module. Would fill structured outcomes for the trials we hold and flag primary-completion dates for the calendar. | Results are posted for a minority of trials and lag by a year or more; outcome measures are free text and need mapping to our endpoint units; sponsor-supplied text can contain third-party copyright. | clinicaltrials.gov | |||||
CTRI (Clinical Trials Registry India) Trials and results Indian registrations, including biosimilar and generic oncology trials rarely registered elsewhere; also indexed by WHO ICTRP. | No API; search forms only; investigator contact details are personal data. | ctri.nic.in | |||||
EU Clinical Trials Register and CTIS Trials and results European registrations and results for studies that never appear on ClinicalTrials.gov, plus EU trial numbers to cross-link. The legacy register is HTML search only; the CTIS public portal has search with CSV export but no documented API and its JSON endpoints refuse scripts. | No stable API; HTML parsing that breaks on redesign. Extracted e-mail addresses must not be used for marketing; results summaries are PDFs. | euclinicaltrials.eu | |||||
ISRCTN registry Trials and results Live XML query API (isrctn.com/api/query) returning full trial records, plus CSV export; the primary registry for UK academic and NIHR-funded trials. The documented API page is currently a 404 but the endpoint works. | Undocumented endpoint could change; pre-2019 metadata has no open licence. | isrctn.com | |||||
jRCT (Japan Registry of Clinical Trials) Trials and results Japanese registrations including many PMDA-facing pivotal studies not on ClinicalTrials.gov; English interface. The registry moved from jrct.niph.go.jp to jrct.mhlw.go.jp. | Explicit anti-scraping notice: link to records, do not ingest. | jrct.mhlw.go.jp | |||||
WHO International Clinical Trials Registry Platform Trials and results One search across 17 primary registries with bridging ids, the only way to de-duplicate a trial registered in several countries. Search results export as XML; the crawling service was unavailable when checked. | Non-commercial licence; no free bulk API; data quality varies by registry. | trialsearch.who.int |
Licences were read from each source's terms page on the date in src/data/data-sources.ts; they change, so check the source before reusing anything. A source being listed here is a statement about its data, not an endorsement, and no source here holds patient records.