PharmKG Query Skill
Search PharmKG triples by entity name. Matching is case-insensitive; exact match
is tried first via a prebuilt index, with substring fallback.
| Input Example |
Matches On |
aspirin |
exact on Entity1_name or Entity2_name |
BRCA1 |
exact / substring on entity names |
Alzheimer Disease |
substring on entity names |
Entity & Relation Types
Entities (~188 k): Drug/Chemical (DrugBank, ChEMBL), Gene/Protein (Entrez, UniProt), Disease (DO, MeSH).
Relations (39 types): chemical–gene (inhibition, activation, binding …), chemical–disease (treatment, marker, risk factor), gene–disease (association, marker), chemical–chemical (similarity, interaction), and others.
API
| Function |
Input |
Returns |
load_pharmkg(path) |
CSV path |
list[dict] of triples |
_build_index(triples) |
triple list |
dict[str, list] (entity→triples) |
search(triples, entity, index=) |
entity string |
list[dict] |
search_batch(triples, entities, index=) |
list of strings |
dict[str, list] |
summarize(hits, entity) |
hits + label |
compact text |
to_json(hits) |
hits |
JSON string |
Usage
See if __name__ == "__main__" block in 55_PharmKG.py for runnable examples
covering: single entity search, batch search, summarize, and JSON output.
Data
- Source:
raw_PharmKG-180k.csv (comma-separated)
- Columns:
Entity1_name, relationship_type, Entity2_name
- Scale: ~188 k entities, 39 relation types, >1 M triples
- Path:
DATA_PATH variable in 55_PharmKG.py
- Paper: Zheng et al., Briefings in Bioinformatics 22(4), 2021. DOI: 10.1093/bib/bbaa344
1---2name: pharmkg-query3description: Query the PharmKG knowledge graph (180k entities, 39 relation types, >1M triples). Use whenever the user asks about biomedical relationships among genes, drugs/chemicals, and diseases — e.g. drug–gene interactions, drug–disease associations, gene–disease links, or drug–drug relationships derived from literature and curated databases.4---5
6# PharmKG Query Skill
7
8Search PharmKG triples by entity name. Matching is case-insensitive; exact match
9is tried first via a prebuilt index, with substring fallback.
10
11| Input Example | Matches On |
12|---|---|
13| `aspirin` | exact on Entity1\_name or Entity2\_name |
14| `BRCA1` | exact / substring on entity names |
15| `Alzheimer Disease` | substring on entity names |
16
17## Entity & Relation Types
18
19**Entities** (~188 k): Drug/Chemical (DrugBank, ChEMBL), Gene/Protein (Entrez, UniProt), Disease (DO, MeSH).
20
21**Relations** (39 types): chemical–gene (inhibition, activation, binding …), chemical–disease (treatment, marker, risk factor), gene–disease (association, marker), chemical–chemical (similarity, interaction), and others.
22
23## API
24
25| Function | Input | Returns |
26|---|---|---|
27| `load_pharmkg(path)` | CSV path | `list[dict]` of triples |
28| `_build_index(triples)` | triple list | `dict[str, list]` (entity→triples) |
29| `search(triples, entity, index=)` | entity string | `list[dict]` |
30| `search_batch(triples, entities, index=)` | list of strings | `dict[str, list]` |
31| `summarize(hits, entity)` | hits + label | compact text |
32| `to_json(hits)` | hits | JSON string |
33
34## Usage
35
36See `if __name__ == "__main__"` block in `55_PharmKG.py` for runnable examples
37covering: single entity search, batch search, summarize, and JSON output.
38
39## Data
40
41- **Source**: `raw_PharmKG-180k.csv` (comma-separated)
42- **Columns**: `Entity1_name`, `relationship_type`, `Entity2_name`
43- **Scale**: ~188 k entities, 39 relation types, >1 M triples
44- **Path**: `DATA_PATH` variable in `55_PharmKG.py`
45- **Paper**: Zheng et al., *Briefings in Bioinformatics* 22(4), 2021. DOI: 10.1093/bib/bbaa344
gabrielmoreira