Update app.py
Browse files
app.py
CHANGED
@@ -1,9 +1,9 @@
|
|
1 |
-
import gradio as gr
|
2 |
import sys
|
3 |
import random
|
4 |
import os
|
5 |
import pandas as pd
|
6 |
import torch
|
|
|
7 |
from torch.utils.data import DataLoader
|
8 |
from transformers import AutoTokenizer
|
9 |
|
@@ -23,6 +23,7 @@ class Config:
|
|
23 |
padding_side = "right"
|
24 |
task = "classification"
|
25 |
sequence_col = "sequence"
|
|
|
26 |
|
27 |
|
28 |
# Assuming 'predict_stability' is your function that predicts protein stability
|
@@ -91,13 +92,14 @@ def predict(cfg, sequence):
|
|
91 |
model = PLTNUM_PreTrainedModel.from_pretrained(cfg.model_path, cfg=cfg)
|
92 |
model.to(cfg.device)
|
93 |
|
|
|
94 |
model.eval()
|
95 |
predictions = []
|
96 |
|
97 |
for inputs, _ in dataloader:
|
98 |
inputs = inputs.to(cfg.device)
|
99 |
with torch.no_grad():
|
100 |
-
with torch.amp.autocast(enabled=cfg.use_amp):
|
101 |
preds = (
|
102 |
torch.sigmoid(model(inputs))
|
103 |
if cfg.task == "classification"
|
@@ -105,88 +107,9 @@ def predict(cfg, sequence):
|
|
105 |
)
|
106 |
predictions += preds.cpu().tolist()
|
107 |
outputs = {}
|
|
|
108 |
outputs["raw prediction values"] = predictions
|
109 |
outputs["binary prediction values"] = [1 if x > 0.5 else 0 for x in predictions]
|
110 |
-
return
|
111 |
|
112 |
-
|
113 |
-
# Gradio Interface
|
114 |
-
with gr.Blocks() as demo:
|
115 |
-
gr.Markdown(
|
116 |
-
"""
|
117 |
-
# PLTNUM: Protein LifeTime Neural Model
|
118 |
-
**Predict the protein half-life from its sequence or PDB file.**
|
119 |
-
"""
|
120 |
-
)
|
121 |
-
|
122 |
-
gr.Image(
|
123 |
-
"https://github.com/sagawatatsuya/PLTNUM/blob/main/model-image.png?raw=true",
|
124 |
-
label="Model Image",
|
125 |
-
)
|
126 |
-
|
127 |
-
# Model and Organism selection in the same row to avoid layout issues
|
128 |
-
with gr.Row():
|
129 |
-
model_choice = gr.Radio(
|
130 |
-
choices=["SaProt", "ESM2"],
|
131 |
-
label="Select PLTNUM's base model.",
|
132 |
-
value="SaProt",
|
133 |
-
)
|
134 |
-
organism_choice = gr.Radio(
|
135 |
-
choices=["Mouse", "Human"],
|
136 |
-
label="Select the target organism.",
|
137 |
-
value="Mouse",
|
138 |
-
)
|
139 |
-
|
140 |
-
with gr.Tabs():
|
141 |
-
with gr.TabItem("Upload PDB File"):
|
142 |
-
gr.Markdown("### Upload your PDB file:")
|
143 |
-
pdb_file = gr.File(label="Upload PDB File")
|
144 |
-
|
145 |
-
predict_button = gr.Button("Predict Stability")
|
146 |
-
prediction_output = gr.Textbox(
|
147 |
-
label="Stability Prediction", interactive=False
|
148 |
-
)
|
149 |
-
|
150 |
-
predict_button.click(
|
151 |
-
fn=predict_stability,
|
152 |
-
inputs=[model_choice, organism_choice, pdb_file],
|
153 |
-
outputs=prediction_output,
|
154 |
-
)
|
155 |
-
|
156 |
-
with gr.TabItem("Enter Protein Sequence"):
|
157 |
-
gr.Markdown("### Enter the protein sequence:")
|
158 |
-
sequence = gr.Textbox(
|
159 |
-
label="Protein Sequence",
|
160 |
-
placeholder="Enter your protein sequence here...",
|
161 |
-
lines=8,
|
162 |
-
)
|
163 |
-
predict_button = gr.Button("Predict Stability")
|
164 |
-
prediction_output = gr.Textbox(
|
165 |
-
label="Stability Prediction", interactive=False
|
166 |
-
)
|
167 |
-
|
168 |
-
predict_button.click(
|
169 |
-
fn=predict_stability,
|
170 |
-
inputs=[model_choice, organism_choice, sequence],
|
171 |
-
outputs=prediction_output,
|
172 |
-
)
|
173 |
-
|
174 |
-
gr.Markdown(
|
175 |
-
"""
|
176 |
-
### How to Use:
|
177 |
-
- **Select Model**: Choose between 'SaProt' or 'ESM2' for your prediction.
|
178 |
-
- **Select Organism**: Choose between 'Mouse' or 'Human'.
|
179 |
-
- **Upload PDB File**: Choose the 'Upload PDB File' tab and upload your file.
|
180 |
-
- **Enter Sequence**: Alternatively, switch to the 'Enter Protein Sequence' tab and input your sequence.
|
181 |
-
- **Predict**: Click 'Predict Stability' to receive the prediction.
|
182 |
-
"""
|
183 |
-
)
|
184 |
-
|
185 |
-
gr.Markdown(
|
186 |
-
"""
|
187 |
-
### About the Tool
|
188 |
-
This tool allows researchers and scientists to predict the stability of proteins using advanced algorithms. It supports both PDB file uploads and direct sequence input.
|
189 |
-
"""
|
190 |
-
)
|
191 |
-
|
192 |
-
demo.launch()
|
|
|
|
|
1 |
import sys
|
2 |
import random
|
3 |
import os
|
4 |
import pandas as pd
|
5 |
import torch
|
6 |
+
import itertools
|
7 |
from torch.utils.data import DataLoader
|
8 |
from transformers import AutoTokenizer
|
9 |
|
|
|
23 |
padding_side = "right"
|
24 |
task = "classification"
|
25 |
sequence_col = "sequence"
|
26 |
+
seed = 42
|
27 |
|
28 |
|
29 |
# Assuming 'predict_stability' is your function that predicts protein stability
|
|
|
92 |
model = PLTNUM_PreTrainedModel.from_pretrained(cfg.model_path, cfg=cfg)
|
93 |
model.to(cfg.device)
|
94 |
|
95 |
+
# predictions = predict_fn(loader, model, cfg)
|
96 |
model.eval()
|
97 |
predictions = []
|
98 |
|
99 |
for inputs, _ in dataloader:
|
100 |
inputs = inputs.to(cfg.device)
|
101 |
with torch.no_grad():
|
102 |
+
with torch.amp.autocast(cfg.device, enabled=cfg.use_amp):
|
103 |
preds = (
|
104 |
torch.sigmoid(model(inputs))
|
105 |
if cfg.task == "classification"
|
|
|
107 |
)
|
108 |
predictions += preds.cpu().tolist()
|
109 |
outputs = {}
|
110 |
+
predictions = list(itertools.chain.from_iterable(predictions))
|
111 |
outputs["raw prediction values"] = predictions
|
112 |
outputs["binary prediction values"] = [1 if x > 0.5 else 0 for x in predictions]
|
113 |
+
return outputs
|
114 |
|
115 |
+
predict_stability("SaProt", "Human", sequence="MELKQK")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|