PMDA provider¶
The Pharmaceuticals and Medical Devices Agency (PMDA) provider contributes Japanese new-drug approval indications. Mira extracts approval rows from PMDA's consolidated PDF and creates regulatory Clinical Reports from their active ingredients and notes.
See PMDA approved products for the source page, expected PDF structure, local path, and format-change checks.
What PMDA contributes¶
Every PMDA report has these values:
| Clinical Report field | Value |
|---|---|
provider |
PMDA |
source |
PMDA |
origin |
REGULATORY_AGENCY |
type |
INDICATION |
clinicalStage |
APPROVAL |
year |
Year extracted from the approval date, when available. |
PMDA does not provide a stable report identifier for the extracted drug–disease pair in this file. Mira creates one by hashing the cleaned drug and extracted disease labels.
Selection and label rules¶
The PDF parser searches for tables containing the expected approval, active-ingredient, and notes headings. It reuses the most recently recognised column layout when a continued table does not repeat its header.
The parser then:
- keeps rows whose approval field contains
Approval; - keeps all data rows when a recognised table has no approval-type column;
- removes rows that contain neither an active ingredient nor notes;
- cleans and separates multiple active ingredients;
- applies disease named-entity recognition (NER) to the notes;
- removes drug–disease rows where either extracted label is empty;
- removes duplicate rows.
Warning
The parser is sensitive to changes in the PDF layout. The checks described on the data-input page should be repeated whenever the source PDF changes.
Mapping behavior¶
The active-ingredient and NER-derived disease labels initially have null identifiers. Disease NER extracts a mention from free text; it does not assign an EFO identifier. The shared entity-mapping stage maps the labels to ChEMBL and EFO after provider reports are combined.