Spaces:
Sleeping
Sleeping
Upload 3 files
Browse files- DNABERT2-FINAL.py +67 -0
- Dockerfile +24 -0
- requirements.txt +9 -0
DNABERT2-FINAL.py
ADDED
@@ -0,0 +1,67 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
|
2 |
+
import streamlit as st
|
3 |
+
from transformers import AutoTokenizer, AutoModelForSequenceClassification, BertConfig
|
4 |
+
import torch
|
5 |
+
|
6 |
+
|
7 |
+
model_name = "rashiqua/dnabert2_epigenetic"
|
8 |
+
|
9 |
+
|
10 |
+
config = BertConfig.from_pretrained(model_name)
|
11 |
+
|
12 |
+
|
13 |
+
tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
|
14 |
+
|
15 |
+
|
16 |
+
model = AutoModelForSequenceClassification.from_pretrained(model_name, trust_remote_code=True, config=config)
|
17 |
+
|
18 |
+
def main():
|
19 |
+
st.title("Epigenetic Marks Prediction")
|
20 |
+
st.write("An application of DNA BERT2")
|
21 |
+
|
22 |
+
|
23 |
+
st.sidebar.header("About")
|
24 |
+
st.sidebar.write("This app uses DNA BERT2 to predict the presence of epigenetic marks in a given DNA sequence.")
|
25 |
+
|
26 |
+
|
27 |
+
user_input = st.text_area("Enter a DNA sequence:", height=150)
|
28 |
+
|
29 |
+
if st.button("Classify Sequence"):
|
30 |
+
if user_input:
|
31 |
+
|
32 |
+
predicted_class, confidence = pred(user_input)
|
33 |
+
|
34 |
+
|
35 |
+
st.subheader("Prediction Result")
|
36 |
+
if predicted_class == 1:
|
37 |
+
st.success("Epigenetic Mark detected!")
|
38 |
+
else:
|
39 |
+
st.info("No epigenetic mark found.")
|
40 |
+
|
41 |
+
st.subheader("Class Distribution")
|
42 |
+
st.write("1 - Epigenetic mark found")
|
43 |
+
st.progress(confidence)
|
44 |
+
st.text(f"{confidence * 100:.2f}%")
|
45 |
+
|
46 |
+
st.write("0 - Epigenetic mark not found")
|
47 |
+
st.progress(1 - confidence)
|
48 |
+
st.text(f"{(1 - confidence) * 100:.2f}%")
|
49 |
+
|
50 |
+
else:
|
51 |
+
st.warning("Please enter a DNA sequence for classification.")
|
52 |
+
|
53 |
+
|
54 |
+
def pred(sequence):
|
55 |
+
encoded_input = tokenizer(sequence, return_tensors='pt')
|
56 |
+
|
57 |
+
with torch.no_grad():
|
58 |
+
outputs = model(input_ids=encoded_input['input_ids'], attention_mask=encoded_input['attention_mask'])
|
59 |
+
logits = outputs[0]
|
60 |
+
predicted_class = logits.argmax(-1).item()
|
61 |
+
confidence = logits.softmax(dim=-1)[0, 1].item()
|
62 |
+
|
63 |
+
return predicted_class, confidence
|
64 |
+
|
65 |
+
|
66 |
+
if __name__ == "__main__":
|
67 |
+
main()
|
Dockerfile
ADDED
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
FROM python:3.12
|
2 |
+
|
3 |
+
RUN useradd -m -u 1000 user
|
4 |
+
|
5 |
+
USER user
|
6 |
+
|
7 |
+
ENV HOME=/home/user \
|
8 |
+
PATH=/home/user/.local/bin:$PATH
|
9 |
+
|
10 |
+
WORKDIR $HOME/app
|
11 |
+
|
12 |
+
RUN pip install --no-cache-dir --upgrade pip
|
13 |
+
|
14 |
+
COPY --chown=user . $HOME/app
|
15 |
+
|
16 |
+
RUN pip install --user -r requirements.txt
|
17 |
+
|
18 |
+
RUN pip uninstall -y triton
|
19 |
+
|
20 |
+
RUN mkdir -p $HOME/.cache/huggingface && chmod 777 $HOME/.cache/huggingface
|
21 |
+
|
22 |
+
EXPOSE 7860
|
23 |
+
|
24 |
+
CMD ["streamlit", "run", "DNABERT2-FINAL.py", "--server.port=7860", "--server.address=0.0.0.0"]
|
requirements.txt
ADDED
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
einops
|
2 |
+
transformers
|
3 |
+
peft
|
4 |
+
omegaconf
|
5 |
+
torch
|
6 |
+
evaluate
|
7 |
+
accelerate
|
8 |
+
streamlit
|
9 |
+
|