app.py

import gradio as gr
import torch
import pandas as pd
from transformers import EsmTokenizer
from model import CleavageSiteModel

# Load tokenizer and model
tokenizer = EsmTokenizer.from_pretrained("tokenizer")  # Path to tokenizer folder
model = CleavageSiteModel(num_classes=75, base_model="facebook/esm2_t30_150M_UR50D")
model.load_state_dict(torch.load("model.pt", map_location="cpu"))
model.eval()

# Load example sequences and labels from CSV
examples_df = pd.read_csv("example_inputs.csv")
examples = examples_df[["sequence", "cleavage_site"]].values.tolist()


# Inference function accepting both sequence and true label
def predict(sequence, true_site):
    inputs = tokenizer(sequence, return_tensors="pt", truncation=True, padding=True)
    with torch.no_grad():
        outputs = model(**inputs)
        logits = outputs["logits"]
        prediction = logits.argmax(dim=1).item()
    return f"Predicted cleavage site index: {prediction} (True: {true_site})"


# Launch Gradio interface
gr.Interface(
    fn=predict,
    inputs=[
        gr.Textbox(label="Protein Sequence", lines=2),
        gr.Number(label="True Cleavage Site")
    ],
    outputs=gr.Textbox(label="Model Output"),
    examples=examples,
    title="Signal Peptide Cleavage Site Predictor",
    description="Created by Nicolai Thorer Sivesind & Erlend Rønning \n Select an example or enter your own protein "
                " sequence and (optionally) its known cleavage site index."
).launch()