f-galkin commited on
Commit
8b4e8fc
1 Parent(s): 8563f22

Added dataset description

Browse files
Files changed (1) hide show
  1. README.md +34 -2
README.md CHANGED
@@ -15,13 +15,14 @@ configs:
15
  data_files: data/batman_formulas.csv
16
  sep: "\t"
17
  task_categories:
18
- - zero-shot-classification
 
19
  language:
20
  - en
21
  - cn
22
  tags:
23
  - biology
24
- - life sciences
25
  - chemistry
26
  - medical
27
  ---
@@ -30,6 +31,37 @@ tags:
30
  This dataset is a reformatted replica of [BATMAN-2.0](http://bionet.ncpsb.org.cn/batman-tcm/#/home), a database of traditional Chinese medicine (TCM) ingredients, herbs, and formulas.
31
  Refer to the original peer-reviewed [publication](https://doi.org/10.1093/nar/gkad926) for more detail.
32
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
33
 
34
  ## Cite this dataset
35
 
 
15
  data_files: data/batman_formulas.csv
16
  sep: "\t"
17
  task_categories:
18
+ - drug-discovery
19
+ - chemical-screening
20
  language:
21
  - en
22
  - cn
23
  tags:
24
  - biology
25
+ - life-sciences
26
  - chemistry
27
  - medical
28
  ---
 
31
  This dataset is a reformatted replica of [BATMAN-2.0](http://bionet.ncpsb.org.cn/batman-tcm/#/home), a database of traditional Chinese medicine (TCM) ingredients, herbs, and formulas.
32
  Refer to the original peer-reviewed [publication](https://doi.org/10.1093/nar/gkad926) for more detail.
33
 
34
+ ## What can I do with this dataset?
35
+ Check out the demo notebooks in the `demo` folder.
36
+ BATMAN is a powerful source of data that links natural compounds and their compositions to protein targets.
37
+ You can use this information to run drug discovery experiments or to extend the functionality of your own biomedical language models.
38
+
39
+ For example, we used this dataset to extend the capabilities of [Precious3GPT](https://doi.org/10.57967/hf/2699) in the `demo/TCM_geroprotectors` demo project.
40
+ While P3GPT is a transformer that can generate omics signatures for various tissues and conditions, it lacks any information on herbal medicine and TCM formulations.
41
+ With BATMAN, we managed to remove this drawback and enable TCM screening in P3GPT.
42
+
43
+ ## Dataset structure
44
+ - `batman_ingredients`: compounds naturally occurring in TCM herbs
45
+ - `UID`: unique ID of an entity throughout the complete dataset;
46
+ - `cid`: [PubChem](https://pubchem.ncbi.nlm.nih.gov/) molecule ID;
47
+ - `pref_name`: conventional name of a compound (as mentioned in BATMAN);
48
+ - `synonyms`: other names associated with the CID (if present in BATMAN);
49
+ - `targets_known`: Verified human protein targets, according to BATMAN. The format is "symbol(ENTREZ_ID)"
50
+ - `targts_pred`: All predicted protein targets, as stated in BATMAN (no significance threshold was used to filter these lists);
51
+ - `herbs`: A list of herbs, in which the compound is encountered;
52
+ - `formulas`: A list of TCM formulas that feature this compound.
53
+ - `batman_herb`: herbs used in TCM
54
+ - `UID`;
55
+ - `pref_name`: Most commonly, the [pinyin](https://en.wikipedia.org/wiki/Pinyin) name of a herb. If not available — any other available name, e.g. latin or common English;
56
+ - `synonyms`
57
+ - `ingredients`: A list of all compound CIDs occurring in a herb;
58
+ - `formulas`: A list of TCM formulas that feature thisherb.
59
+ - `batman_formulas`: TCM formulas featured in the BATMAN dataset;
60
+ - `UID`;
61
+ - `pref_name`: The pinyin name of a herbal drug. Non-identical compositions sharing the same name are disambiguated using the `~X` suffix;
62
+ - `synonyms`: The Chinese hieroglyphic spelling of the compostion;
63
+ - `ingredients`;
64
+ - `herbs`.
65
 
66
  ## Cite this dataset
67