Datasets:

DOI:
License:
rcalef's picture
Add instruction tuning datasets (#3)
36fe3c0 verified
|
raw
history blame
762 Bytes
`disgenet_info_filtered.pkl` is a pickled pandas DataFrame containing disease descriptions sourced from:
- UMLS
- OrphaNet (via PrimeKG)
- MONDO (via PrimeKG)
Columns are:
- index - unique identifier for this disease / UMLS concept ID in the txPLM datasets
- disgenet_id - UMLS ID
- various columns named by the sources of the description
- allDescriptions - concatenation of all descriptions in the other columns
- numDescriptions - total number of unique free text descriptions for this concept ID
Descriptions are concatenated with the `|` symbol, both within one source in the case of multiple
descriptions from one source, or across all sources in the `allDescriptions` column.
For more details, see the notebook at `../disgenet_curation.ipynb`.