The FDA regulatory approvals slot is currently defined as "Numbers that identify specific drug applications", an FDA-scoped concept. In practice, producers populate it with application-number values that are not guaranteed to be FDA identifiers (unregistered numbers, label-sourced IDs with invalid or missing application-type prefixes, bare digit strings carried as provenance). This creates a semantic mismatch between the slot's definition and its deployed usage.
Current deployed misuse
Already-deployed values in this field do include non-FDA approvals; this is not a hypothetical risk but the current state of published data:
- The Translator ingest for the Drug Approvals KP (
NCATSTranslator/translator-ingests, src/translator_ingest/ingests/dakp/dakp.py) maps the upstream approvals column directly to FDA_regulatory_approvals with no validation that values are FDA application numbers.
- The upstream pipeline emits unresolvable values by design, treating the field as general application-number provenance rather than verified FDA identifiers. Values currently deployed on edges include bare digit strings absent from any FDA register and label-sourced IDs carrying invalid application-type prefixes.
Proposed resolutions (maintainer decision requested)
- Broaden the existing slot: relax the
FDA regulatory approvals description to cover drug-application provenance identifiers generally, not only verified FDA applications.
- Add an agency-agnostic attribute (e.g.,
regulatory approvals) for general application-number provenance; keep or remove FDA_regulatory_approvals depending on how we feel.
Both are acceptable; option 1 is lower-friction, option 2 preserves strict semantics for existing consumers.
Impact
No model change breaks current data on its own; resolution 2 would require a migration path for producers currently emitting non-FDA values under the FDA-scoped slot. Note that we intend to add European drug approvals (EMA marketing authorizations) in the future, which strengthens the case for an agency-agnostic home for these values.
The
FDA regulatory approvalsslot is currently defined as "Numbers that identify specific drug applications", an FDA-scoped concept. In practice, producers populate it with application-number values that are not guaranteed to be FDA identifiers (unregistered numbers, label-sourced IDs with invalid or missing application-type prefixes, bare digit strings carried as provenance). This creates a semantic mismatch between the slot's definition and its deployed usage.Current deployed misuse
Already-deployed values in this field do include non-FDA approvals; this is not a hypothetical risk but the current state of published data:
NCATSTranslator/translator-ingests,src/translator_ingest/ingests/dakp/dakp.py) maps the upstreamapprovalscolumn directly toFDA_regulatory_approvalswith no validation that values are FDA application numbers.Proposed resolutions (maintainer decision requested)
FDA regulatory approvalsdescription to cover drug-application provenance identifiers generally, not only verified FDA applications.regulatory approvals) for general application-number provenance; keep or removeFDA_regulatory_approvalsdepending on how we feel.Both are acceptable; option 1 is lower-friction, option 2 preserves strict semantics for existing consumers.
Impact
No model change breaks current data on its own; resolution 2 would require a migration path for producers currently emitting non-FDA values under the FDA-scoped slot. Note that we intend to add European drug approvals (EMA marketing authorizations) in the future, which strengthens the case for an agency-agnostic home for these values.