Datasets:

Modalities:
Text
Formats:
parquet
Libraries:
Datasets
Dask
License:
andrecornman commited on
Commit
9fdf3ee
·
verified ·
1 Parent(s): 9ff3657

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +39 -0
README.md CHANGED
@@ -16,4 +16,43 @@ configs:
16
  data_files:
17
  - split: train
18
  path: data/train-*
 
19
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
16
  data_files:
17
  - split: train
18
  path: data/train-*
19
+ license: cc-by-sa-4.0
20
  ---
21
+
22
+ # OG_prot90: An Open Genomic Protein Dataset
23
+ The `OG_prot90` dataset is a protein-only dataset, created by clustering the Open Genomic dataset ([`OG`](https://huggingface.co/datasets/tattabio/OG)) at 90% sequence identity.
24
+
25
+ MMseqs2 linclust (Steinegger and Söding 2018) was used to cluster all 400M protein sequences from the OG dataset, resulting in 85M protein sequences.
26
+ Sequences were clustered at 90% sequence id and 90% sequence coverage.
27
+
28
+
29
+ ## Use
30
+
31
+ ```python
32
+ import datasets
33
+ ds = datasets.load_dataset('tattabio/OG_prot90')
34
+ ```
35
+
36
+ To preview the dataset without downloading, load in streaming mode:
37
+ ```python
38
+ import datasets
39
+ ds = datasets.load_dataset('tattabio/OG_prot90', streaming=True)['train']
40
+ print(next(iter(ds)))
41
+ ```
42
+
43
+
44
+
45
+ ## Citation
46
+
47
+ **BibTeX:**
48
+
49
+ ```
50
+ @article{Cornman2024,
51
+ title = {The OMG dataset: An Open MetaGenomic corpus for mixed-modality genomic language modeling},
52
+ url = {https://www.biorxiv.org/content/early/2024/08/17/2024.08.14.607850},
53
+ DOI = {10.1101/2024.08.14.607850},
54
+ publisher = {Cold Spring Harbor Laboratory},
55
+ author = {Cornman, Andre and West-Roberts, Jacob and Camargo, Antonio Pedro and Roux, Simon and Beracochea, Martin and Mirdita, Milot and Ovchinnikov, Sergey and Hwang, Yunha},
56
+ year = {2024},
57
+ }
58
+ ```