Update README.md
Browse files
README.md
CHANGED
@@ -3,4 +3,31 @@ license: mit
|
|
3 |
tags:
|
4 |
- protein
|
5 |
- thermostability
|
6 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
3 |
tags:
|
4 |
- protein
|
5 |
- thermostability
|
6 |
+
---
|
7 |
+
|
8 |
+
__Purpose__: classifies protein sequence into Thermophilic (> 60C) or Mesophilic (<40C) by host organism growth temperature.
|
9 |
+
|
10 |
+
__Training__:
|
11 |
+
ProteinBERT (Rostlab/prot_bert) was fine tuned on a class balanced version of learn2therm (see [here]()), about 250k protein amino acid sequences.
|
12 |
+
|
13 |
+
Training parameters below:
|
14 |
+
TODO
|
15 |
+
|
16 |
+
See the [training repository](https://github.com/BeckResearchLab/learn2thermML) for code.
|
17 |
+
|
18 |
+
__Usage__:
|
19 |
+
Prepare sequences identically to using the original pretrained model:
|
20 |
+
|
21 |
+
```
|
22 |
+
from transformers import BertModelForSequenceClassification, BertTokenizer
|
23 |
+
import torch
|
24 |
+
import re
|
25 |
+
tokenizer = BertTokenizer.from_pretrained("evankomp/learn2therm", do_lower_case=False )
|
26 |
+
model = BertModelForSequenceClassification.from_pretrained("evankomp/learn2therm")
|
27 |
+
sequence_Example = "A E T C Z A O"
|
28 |
+
sequence_Example = re.sub(r"[UZOB]", "X", sequence_Example)
|
29 |
+
encoded_input = tokenizer(sequence_Example, return_tensors='pt')
|
30 |
+
output = torch.argmax(model(**encoded_input), dim=1)
|
31 |
+
```
|
32 |
+
|
33 |
+
1 indicates thermophilic, 0 mesophilic.
|