drugprot

Name: drugprot
Author: QSong-github

132stars4forks

How to install drugprot?

qsong-github/drugclaw/drugprot

$npx -y skills add QSong-github/DrugClaw --skill drugprot

Installs into the current project.

Summary

Query drug/chemical and gene/protein entities in the BioCreative VII DrugProt dataset. Returns annotated relations (e.g., INHIBITOR, ACTIVATOR, SUBSTRATE) between chemicals and genes/proteins from biomedical literature.

Files · 1

View on GitHub

SKILL.md

1	# 31_DrugProt — Drug-Protein Relation Query
2
3	## Overview
4
5	Query drug/chemical and gene/protein entities in the BioCreative VII DrugProt dataset. Returns annotated relations (e.g., INHIBITOR, ACTIVATOR, SUBSTRATE) between chemicals and genes/proteins from biomedical literature.
6
7	- Source: <https://zenodo.org/records/5119892>
8	- Paper: <https://pmc.ncbi.nlm.nih.gov/articles/PMC10683943/>
9	- Coverage: 15,000 PubMed abstracts (3,500 training + 750 development + 10,750 test-background), 13 relation types
10
11	## Data Location
12
13	```
14	DrugClaw/
15	├── skills/drug_nlp/drugprot/
16	│ └── example.py ← helper script
17	└── resources_metadata/drug_nlp/DrugProt/
18	└── drugprot-gs-training-development/
19	├── training/
20	│ ├── drugprot_training_abstracs.tsv
21	│ ├── drugprot_training_entities.tsv
22	│ └── drugprot_training_relations.tsv
23	├── development/
24	│ ├── drugprot_development_abstracs.tsv
25	│ ├── drugprot_development_entities.tsv
26	│ └── drugprot_development_relations.tsv
27	└── test-background/
28	├── test_background_abstracts.tsv
29	└── test_background_entities.tsv
30	```
31
32	Use the repo-local `resources_metadata/drug_nlp/DrugProt/drugprot-gs-training-development` path, or override it with the `DRUGPROT_DIR` environment variable if your dataset lives elsewhere.
33
34	All three splits are loaded by default. Note: `test-background` has no relations file (relations are the prediction target).
35
36	## Quick Start
37
38	```python
39	from importlib.util import spec_from_file_location, module_from_spec
40
41	# Load module
42	spec = spec_from_file_location("drugprot", "/path/to/skills/drug_nlp/drugprot/example.py")
43	dp = module_from_spec(spec)
44	spec.loader.exec_module(dp)
45
46	# Load dataset (one-time, ~2 s)
47	ds = dp.load_dataset("/path/to/drugprot-gs-training-development")
48
49	# Query single entity
50	results = dp.query_entities(ds, "aspirin")
51	print(dp.format_results(results))
52
53	# Query multiple entities
54	results = dp.query_entities(ds, ["metformin", "insulin", "EGFR"])
55	print(dp.format_results(results))
56	```
57
58	## API
59
60	### `load_dataset(base_dir, splits=["training","development","test-background"]) -> dict`
61
62	Loads and indexes all TSV files. Returns a dict with keys: `abstracts`, `entities`, `relations`, `name_index`.
63
64	### `query_entities(dataset, names, case_sensitive=False) -> list[dict]`
65
66	\| Parameter \| Type \| Description \|
67	\|---\|---\|---\|
68	\| `dataset` \| `dict` \| Output of `load_dataset()` \|
69	\| `names` \| `str` or `list[str]` \| Entity name(s) to query \|
70	\| `case_sensitive` \| `bool` \| Default `False`; falls back to substring match if exact match fails \|
71
72	Returns a list of result dicts, one per query name:
73
74	```json
75	[
76	{
77	"query": "aspirin",
78	"matches": [
79	{
80	"pmid": "12345678",
81	"entity_id": "T3",
82	"entity_type": "CHEMICAL",
83	"entity_text": "aspirin",
84	"relations": [
85	{
86	"relation_type": "INHIBITOR",
87	"partner_id": "T12",
88	"partner_text": "COX-2",
89	"partner_type": "GENE-Y",
90	"role": "arg1(chemical)"
91	}
92	],
93	"article_title": "Effects of aspirin on ..."
94	}
95	]
96	}
97	]
98	```
99
100	### `format_results(results, max_matches=5) -> str`
101
102	Formats query results into a concise, LLM-readable plain-text summary.
103
104	## Relation Types
105
106	\| Type \| Description \|
107	\|---\|---\|
108	\| INHIBITOR \| Chemical inhibits gene/protein \|
109	\| ACTIVATOR \| Chemical activates gene/protein \|
110	\| AGONIST \| Chemical is an agonist \|
111	\| ANTAGONIST \| Chemical is an antagonist \|
112	\| SUBSTRATE \| Chemical is a substrate \|
113	\| PRODUCT-OF \| Chemical is a product of enzyme \|
114	\| INDIRECT-DOWNREGULATOR \| Chemical indirectly downregulates \|
115	\| INDIRECT-UPREGULATOR \| Chemical indirectly upregulates \|
116	\| DIRECT-REGULATOR \| Chemical directly regulates \|
117	\| PART-OF \| Chemical is part of protein complex \|
118	\| COFACTOR \| Chemical acts as cofactor \|
119	\| NOT \| Negative relation \|
120	\| UNDEFINED \| Undefined relation \|
121
122	## CLI Usage
123
124	```bash
125	export DRUGPROT_DIR=/path/to/drugprot-gs-training-development
126	python 31_DrugProt.py aspirin insulin p53
127	```

People who liked this also viewed

6 picks

Type

drugprot

132stars4forks

View on GitHub

How to install drugprot?

qsong-github/drugclaw/drugprot

$npx -y skills add QSong-github/DrugClaw --skill drugprot

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# 31_DrugProt — Drug-Protein Relation Query
2
3	## Overview
4
5	Query drug/chemical and gene/protein entities in the BioCreative VII DrugProt dataset. Returns annotated relations (e.g., INHIBITOR, ACTIVATOR, SUBSTRATE) between chemicals and genes/proteins from biomedical literature.
6
7	- Source: <https://zenodo.org/records/5119892>
8	- Paper: <https://pmc.ncbi.nlm.nih.gov/articles/PMC10683943/>
9	- Coverage: 15,000 PubMed abstracts (3,500 training + 750 development + 10,750 test-background), 13 relation types
10
11	## Data Location
12
13	```
14	DrugClaw/
15	├── skills/drug_nlp/drugprot/
16	│ └── example.py ← helper script
17	└── resources_metadata/drug_nlp/DrugProt/
18	└── drugprot-gs-training-development/
19	├── training/
20	│ ├── drugprot_training_abstracs.tsv
21	│ ├── drugprot_training_entities.tsv
22	│ └── drugprot_training_relations.tsv
23	├── development/
24	│ ├── drugprot_development_abstracs.tsv
25	│ ├── drugprot_development_entities.tsv
26	│ └── drugprot_development_relations.tsv
27	└── test-background/
28	├── test_background_abstracts.tsv
29	└── test_background_entities.tsv
30	```
31
32	Use the repo-local `resources_metadata/drug_nlp/DrugProt/drugprot-gs-training-development` path, or override it with the `DRUGPROT_DIR` environment variable if your dataset lives elsewhere.
33
34	All three splits are loaded by default. Note: `test-background` has no relations file (relations are the prediction target).
35
36	## Quick Start
37
38	```python
39	from importlib.util import spec_from_file_location, module_from_spec
40
41	# Load module
42	spec = spec_from_file_location("drugprot", "/path/to/skills/drug_nlp/drugprot/example.py")
43	dp = module_from_spec(spec)
44	spec.loader.exec_module(dp)
45
46	# Load dataset (one-time, ~2 s)
47	ds = dp.load_dataset("/path/to/drugprot-gs-training-development")
48
49	# Query single entity
50	results = dp.query_entities(ds, "aspirin")
51	print(dp.format_results(results))
52
53	# Query multiple entities
54	results = dp.query_entities(ds, ["metformin", "insulin", "EGFR"])
55	print(dp.format_results(results))
56	```
57
58	## API
59
60	### `load_dataset(base_dir, splits=["training","development","test-background"]) -> dict`
61
62	Loads and indexes all TSV files. Returns a dict with keys: `abstracts`, `entities`, `relations`, `name_index`.
63
64	### `query_entities(dataset, names, case_sensitive=False) -> list[dict]`
65
66	\| Parameter \| Type \| Description \|
67	\|---\|---\|---\|
68	\| `dataset` \| `dict` \| Output of `load_dataset()` \|
69	\| `names` \| `str` or `list[str]` \| Entity name(s) to query \|
70	\| `case_sensitive` \| `bool` \| Default `False`; falls back to substring match if exact match fails \|
71
72	Returns a list of result dicts, one per query name:
73
74	```json
75	[
76	{
77	"query": "aspirin",
78	"matches": [
79	{
80	"pmid": "12345678",
81	"entity_id": "T3",
82	"entity_type": "CHEMICAL",
83	"entity_text": "aspirin",
84	"relations": [
85	{
86	"relation_type": "INHIBITOR",
87	"partner_id": "T12",
88	"partner_text": "COX-2",
89	"partner_type": "GENE-Y",
90	"role": "arg1(chemical)"
91	}
92	],
93	"article_title": "Effects of aspirin on ..."
94	}
95	]
96	}
97	]
98	```
99
100	### `format_results(results, max_matches=5) -> str`
101
102	Formats query results into a concise, LLM-readable plain-text summary.
103
104	## Relation Types
105
106	\| Type \| Description \|
107	\|---\|---\|
108	\| INHIBITOR \| Chemical inhibits gene/protein \|
109	\| ACTIVATOR \| Chemical activates gene/protein \|
110	\| AGONIST \| Chemical is an agonist \|
111	\| ANTAGONIST \| Chemical is an antagonist \|
112	\| SUBSTRATE \| Chemical is a substrate \|
113	\| PRODUCT-OF \| Chemical is a product of enzyme \|
114	\| INDIRECT-DOWNREGULATOR \| Chemical indirectly downregulates \|
115	\| INDIRECT-UPREGULATOR \| Chemical indirectly upregulates \|
116	\| DIRECT-REGULATOR \| Chemical directly regulates \|
117	\| PART-OF \| Chemical is part of protein complex \|
118	\| COFACTOR \| Chemical acts as cofactor \|
119	\| NOT \| Negative relation \|
120	\| UNDEFINED \| Undefined relation \|
121
122	## CLI Usage
123
124	```bash
125	export DRUGPROT_DIR=/path/to/drugprot-gs-training-development
126	python 31_DrugProt.py aspirin insulin p53
127	```

People who liked this also viewed

6 picks

Type