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

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

# Selected features used in training
selected_features = [
    "A", "R", "N", "D", "C", "E", "Q", "G", "H", "I", "L", "K", "M", "F", "P", "S", "T", "W", "Y", "V",
    "AA", "AR", "AN", "AD", "AC", "AE", "AQ", "AG", "AI", "AL", "AK", "AF", "AP", "AS", "AT", "AY", "AV",
    "RA", "RR", "RN", "RD", "RC", "RE", "RQ", "RG", "RH", "RI", "RL", "RK", "RM", "RF", "RS", "RT", "RY", "RV",
    "NA", "NR", "ND", "NC", "NE", "NG", "NI", "NL", "NK", "NP",
    "DA", "DR", "DN", "DD", "DC", "DE", "DQ", "DG", "DI", "DL", "DK", "DP", "DS", "DT", "DV",
    "CA", "CR", "CN", "CD", "CC", "CE", "CG", "CH", "CI", "CL", "CK", "CF", "CP", "CS", "CT", "CY", "CV",
    "EA", "ER", "EN", "ED", "EC", "EE", "EQ", "EG", "EI", "EL", "EK", "EP", "ES", "ET", "EV",
    "QA", "QR", "QC", "QG", "QL", "QK", "QP", "QT", "QV",
    "GA", "GR", "GD", "GC", "GE", "GQ", "GG", "GI", "GL", "GK", "GF", "GP", "GS", "GW", "GY", "GV",
    "HC", "HG", "HL", "HK", "HP",
    "IA", "IR", "ID", "IC", "IE", "II", "IL", "IK", "IF", "IP", "IS", "IT", "IV",
    "LA", "LR", "LN", "LD", "LC", "LE", "LQ", "LG", "LI", "LL", "LK", "LM", "LF", "LP", "LS", "LT", "LV",
    "KA", "KR", "KN", "KD", "KC", "KE", "KQ", "KG", "KH", "KI", "KL", "KK", "KM", "KF", "KP", "KS", "KT", "KV",
    "MA", "ME", "MI", "ML", "MK", "MF", "MP", "MS", "MT", "MV",
    "FR", "FC", "FQ", "FG", "FI", "FL", "FF", "FS", "FT", "FY", "FV",
    "PA", "PR", "PD", "PC", "PE", "PG", "PL", "PK", "PS", "PV",
    "SA", "SR", "SD", "SC", "SE", "SG", "SH", "SI", "SL", "SK", "SF", "SP", "SS", "ST", "SY", "SV",
    "TA", "TR", "TN", "TC", "TE", "TG", "TI", "TL", "TK", "TF", "TP", "TS", "TT", "TV",
    "WC",
    "YR", "YD", "YC", "YG", "YL", "YS", "YV",
    "VA", "VR", "VD", "VC", "VE", "VQ", "VG", "VI", "VL", "VK", "VP", "VS", "VT", "VY", "VV"
]

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

    # Select features that match training data
    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])
    selected_feature_array = selected_feature_df.T.to_numpy()

    return selected_feature_array

def predict(sequence):
    """Predict if the sequence is an AMP or not."""
    features = extract_features(sequence)
    prediction = model.predict(features.T)[0]
    probabilities = model.predict_proba(features.T)[0]  
    prob_amp = probabilities[0]  
    prob_non_amp = probabilities[1]  

    return f"⚡ {prob_amp * 100:.2f}% chance of being an Antimicrobial Peptide (AMP)" if prediction == 0 else f"❌ {prob_non_amp * 100:.2f}% chance of being Non-AMP"

# Gradio interface
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 whether it's an antimicrobial peptide (AMP) or not."
)

iface.launch(share=True)