app.py

import gradio as gr
import joblib
import numpy as np
import pandas as pd
from propy import AAComposition
from sklearn.preprocessing import MinMaxScaler
from fastapi import FastAPI
from gradio.routes import mount_gradio_app

# Load trained SVM model and scaler
model = joblib.load("SVM.joblib")
scaler = joblib.load("norm.joblib")

# FastAPI instance
app = FastAPI()

# Selected features used in the model
selected_features = [
    "A", "R", "N", "D", "C", "E", "Q", "G", "H", "I", "L", "K", "M", "F", "P", "S", "T", "W", "Y", "V"
]  # (Shortened for brevity)

def extract_features(sequence):
    """Extract only the required features and normalize them."""
    all_features = AAComposition.CalculateAADipeptideComposition(sequence)
    feature_values = list(all_features.values())
    feature_array = np.array(feature_values).reshape(-1, 1)[:420]  # Ensure correct shape
    normalized_features = scaler.transform(feature_array.T).flatten()

    selected_feature_dict = {
        feature: normalized_features[i]
        for i, feature in enumerate(selected_features)
        if feature in all_features
    }

    selected_feature_df = pd.DataFrame([selected_feature_dict])
    return selected_feature_df.to_numpy()

def predict(sequence):
    """Predict AMP vs Non-AMP"""
    features = extract_features(sequence)
    prediction = model.predict(features.T)[0]
    return {"prediction": "AMP" if prediction == 0 else "Non-AMP"}

# FastAPI endpoint
@app.post("/predict/")
async def predict_api(request: dict):
    sequence = request.get("sequence", "")
    if not sequence or len(sequence) < 10 or len(sequence) > 100:
        return {"error": "Sequence length must be between 10 and 100."}
    return predict(sequence)

# Gradio Interface (optional if you want UI access)
iface = gr.Interface(
    fn=predict,
    inputs=gr.Textbox(label="Enter Protein Sequence"),
    outputs=gr.Label(label="Prediction"),
    title="AMP Classifier",
    description="Enter an amino acid sequence to predict AMP or Non-AMP."
)

app = mount_gradio_app(app, iface, path="/gradio")

# Run the server with: `uvicorn filename:app --host 0.0.0.0 --port 7860`