Data Sources

This schedule is the list Terms of Service Section 4.3 points at. It records where each source is acquired, and the licence that source publishes where that page was read.

Training Motif’s own models uses the consent switch, allow_llm_training on the user and on the organization. Motif Dataset is the output Motif generates: the structured findings and the model-written association sentence. The article corpus is not that product.

Literature

Source	How acquired
Cache	Earlier fetch reused before a new download
Europe PMC	Europe PMC API, including preprint full text (`europepmc_preprint`)
PMC	NCBI PMC full text (`pmc_bioc`)
Unpaywall	Unpaywall `is_oa`, then the PDF URL Unpaywall returns
Direct PDF	A `fulltext_url` whose host is on the download allowlist, stored as `fulltext_source` `pdf`. The allowlist includes subscription publishers. This path does not check `is_oa`
Upload	A file a user submits, stored as `fulltext_source` `user_upload`

Showing the text to the organization that requested it is separate from training and from Motif Dataset.

Reference databases

Responses are stored on knowledge.global_cross_references and shown in the workspace of the organization that ran the cross-reference. They are not a training corpus, and they are not Motif Dataset.

Source	How acquired	Licence
Antibody Registry	`https://scicrunch.org/api/1`	RRIDs and core fields (vendor, catalog number, citation) are CC0. Additional vendor metadata is CC-NC. The code is Apache 2.0
ATC	`https://rxnav.nlm.nih.gov/REST`	WHO Collaborating Centre: citation required. Commercial copying and distribution are not allowed. Changing the material is not allowed. RxNav says its API needs no licence except proprietary source vocabularies
BioPortal	`https://data.bioontology.org`	Each ontology keeps its own licence. Where a licensing attribute is set, reuse is the user’s responsibility. No grant over the repository
cBioPortal	`https://www.cbioportal.org/api`	ODC Open Database License. Attribution to the original study. Some studies restrict commercial use
Cell Ontology	`https://www.ebi.ac.uk/ols4/api/ontologies/cl`	CC BY 4.0. Attribution required
Cellosaurus	`https://api.cellosaurus.org`	CC BY 4.0. Attribution required
ChEBI	`https://www.ebi.ac.uk/chebi/backend/api/public`	CC BY 4.0. Attribution required
ChEMBL	`https://www.ebi.ac.uk`	CC BY-SA
ChEMBL ATC	`https://www.ebi.ac.uk`	CC BY-SA
CIViC	`https://civicdb.org/api/graphql`	CC0
ClinicalTrials.gov	`https://clinicaltrials.gov/api/v2/studies`	US government public information
ClinVar	`https://eutils.ncbi.nlm.nih.gov/entrez/eutils`	Submissions are available for unrestricted distribution. Attribution is requested. NCBI places no restriction of its own and cannot license a submitter’s patent or copyright claim
COSMIC signatures	`https://cancer.sanger.ac.uk`	Paid commercial licence. Academic terms forbid redistribution
dbSNP	`https://eutils.ncbi.nlm.nih.gov/entrez/eutils`	NLM public-domain text does not clear submitter records
DGIdb	`https://dgidb.org/api`	MIT is the code. The aggregate has no single data licence. Each source keeps the licence listed at dgidb.org/browse/sources
DrugBank	No DrugBank client. Identifiers are read from UniProt records (`drugbank_ids`)	Non-commercial. No redistribution
EBI OLS	`https://www.ebi.ac.uk/ols4/api`	Apache 2.0 is the software. Each ontology keeps its own licence. EMBL-EBI adds no restriction beyond the data owner and expects attribution
Ensembl	`https://rest.ensembl.org`	Project-generated data has no restrictions. Third-party annotations are not cleared
Ensembl Regulatory	`https://rest.ensembl.org`	Same as Ensembl
ENA taxonomy	`https://www.ebi.ac.uk`	The taxonomy is NCBI’s. EMBL-EBI adds no restriction beyond the data owner and expects attribution. NLM public-domain text does not clear submitter records
FDA	`https://api.fda.gov`	CC0. Some third-party works are not public domain and will be marked
FDA CDx	FDA in-vitro diagnostics pages	fda.gov contents are public domain unless otherwise noted. Credit is appreciated and not required
Gene Ontology	`https://www.ebi.ac.uk/QuickGO/services`	CC BY 4.0. Attribution required
GlycoEpitope	`https://glycoepitope.jp`	Copyright Ritsumeikan University, all rights reserved. No reuse licence is published
GlyGen	`https://api.glygen.org`	Data CC BY 4.0. Attribution required. The code is GPL-3
GlyTouCan	`https://ts.glytoucan.org`	CC0
gnomAD	`https://gnomad.broadinstitute.org`	CC0. No re-identification
GtoPdb	`https://www.guidetopharmacology.org`	CC BY-SA and ODbL
HGNC	`https://rest.genenames.org`	CC0
Human Protein Atlas	`https://www.proteinatlas.org`	CC BY 4.0. Attribution required
IEDB	`https://query-api.iedb.org`	CC BY 4.0. Attribution required
ImmPort	`https://www.immport.org`	Use and distribution for a legal purpose. Redistribution on the same terms. The data provider remains the owner
InnateDB	`https://psicquic.all.innatedb.com/webservices/current/search`	Curated interactions are under the Design Science License. All other data stays under the originating database
IPD-IMGT/HLA	`https://www.ebi.ac.uk/cgi-bin/ipd/api`	CC BY-ND. Commercial use of the unmodified database with citation. A modified copy needs permission
IPD-MHC	`https://www.ebi.ac.uk/cgi-bin/ipd/api`	CC BY-ND. Commercial use of the unmodified database with citation. A modified copy needs permission
iPTMnet	`https://research.bioinformatics.udel.edu`	CC BY-NC-SA 4.0. Not licensed for commercial use
KEGG	No KEGG client. Pathway names arrive inside STRING responses	Paid commercial licence
LIPID MAPS	`https://www.lipidmaps.org/databases/lmsd`	CC BY 4.0. Attribution required
MedGen	`https://eutils.ncbi.nlm.nih.gov/entrez/eutils`	NLM public-domain text does not clear submitter records
MGI	`https://www.mousemine.org/mousemine/service`	CC BY 4.0. Attribution required. Images are not included
MGI strains	`https://www.informatics.jax.org/downloads/reports/MGI_Strain.rpt` and `https://www.findmice.org/report.json`	Same as MGI
Monarch	`https://api.monarchinitiative.org/v3/api`	The software is BSD-3. The graph stays under the upstream licences. No grant over the integrated graph
MSigDB	`https://mygeneset.info`	Releases after April 2017 are CC BY 4.0. Attribution required. KEGG-derived sets are not included
NCBI Gene	`https://eutils.ncbi.nlm.nih.gov`	NLM public-domain text does not clear submitter records
NCBI Taxonomy	`https://api.ncbi.nlm.nih.gov`	NLM public-domain text does not clear submitter records
NCBI Virus	`https://eutils.ncbi.nlm.nih.gov`	NLM public-domain text does not clear submitter records
Open Targets	`https://api.platform.opentargets.org/api/v4/graphql`	CC0 for Open Targets scores. ChEMBL content inside an Open Targets payload is not included
PDCM Finder	`https://www.cancermodels.org/api`	Apache 2.0 is the code. Model metadata is under EMBL-EBI terms of use. Suppliers keep their own terms
PharmGKB	`https://api.clinpgx.org/v1`	CC BY-SA. The publisher also says the data cannot be sold
RCSB PDB	`https://data.rcsb.org/rest/v1/core/entry`	CC0. Molecule of the Month text is separate and is not included
Reactome	`https://reactome.org/ContentService`	CC BY 4.0. Attribution required
RGD	`https://rest.rgd.mcw.edu/rgdws`	Freely provided, with no login and no licence requirement. Not a Creative Commons or ODbL grant. The code is GPL-3
RNAcentral	`https://rnacentral.org/api/v1`	CC0 from version 20
STRING	`https://string-db.org/api`	CC BY 4.0. Attribution required
UniProt	`https://rest.uniprot.org`	CC BY 4.0. Attribution required
VDJdb	`https://raw.githubusercontent.com/antigenomics/vdjdb-db/master/latest-version.txt`	The 2018 paper says the dataset is CC BY-ND 4.0. The GitHub repository’s current licence is AGPL-3.0

Binder design and mRNA design call Motif compute endpoints. They are not reference databases.