Create app.py

app.py

@@ -0,0 +1,212 @@
+import gradio as gr
+import requests
+from rdkit import Chem
+from rdkit.Chem import Descriptors
+import pandas as pd
+from inference import DDIPredictor
+import torch
+import re
+
+# Initialize your trained model
+predictor = DDIPredictor(".")
+
+def fetch_pubchem_data(drug_name):
+    """Fetch drug data from PubChem by name"""
+    try:
+        # First, search for the compound ID
+        search_url = f"https://pubchem.ncbi.nlm.nih.gov/rest/pug/compound/name/{drug_name}/cids/JSON"
+        search_response = requests.get(search_url)
+        
+        if search_response.status_code != 200:
+            return None, f"Drug '{drug_name}' not found in PubChem"
+        
+        cid = search_response.json()['IdentifierList']['CID'][0]
+        
+        # Fetch compound data
+        compound_url = f"https://pubchem.ncbi.nlm.nih.gov/rest/pug/compound/cid/{cid}/property/CanonicalSMILES,MolecularWeight,XLogP,TPSA/JSON"
+        compound_response = requests.get(compound_url)
+        
+        if compound_response.status_code != 200:
+            return None, "Failed to fetch compound data"
+        
+        data = compound_response.json()['PropertyTable']['Properties'][0]
+        
+        # Calculate additional properties from SMILES
+        mol = Chem.MolFromSmiles(data['CanonicalSMILES'])
+        if mol:
+            data['RotatableBondCount'] = Descriptors.NumRotatableBonds(mol)
+            data['HBondDonorCount'] = Descriptors.NumHDonors(mol)
+            data['HBondAcceptorCount'] = Descriptors.NumHAcceptors(mol)
+            data['Complexity'] = Descriptors.MolWt(mol)  # Simplified complexity
+        
+        return data, None
+        
+    except Exception as e:
+        return None, f"Error fetching data: {str(e)}"
+
+def generate_interaction_description(drug1_data, drug2_data):
+    """Generate a clinical-style interaction description based on molecular properties"""
+    
+    descriptions = []
+    
+    # Based on molecular weight difference
+    mw_diff = abs(drug1_data['MolecularWeight'] - drug2_data['MolecularWeight'])
+    if mw_diff > 500:
+        descriptions.append("Significant molecular size difference may affect metabolism")
+    
+    # Based on lipophilicity (XLogP)
+    logp_diff = abs(drug1_data.get('XLogP', 0) - drug2_data.get('XLogP', 0))
+    if logp_diff > 3:
+        descriptions.append("Differing lipophilicity may influence distribution and clearance")
+    
+    # Based on TPSA (polar surface area)
+    tpsa_diff = abs(drug1_data.get('TPSA', 0) - drug2_data.get('TPSA', 0))
+    if tpsa_diff > 100:
+        descriptions.append("Varying polar surface areas suggest different membrane permeability")
+    
+    # Based on hydrogen bonding
+    hbond_total1 = drug1_data.get('HBondDonorCount', 0) + drug1_data.get('HBondAcceptorCount', 0)
+    hbond_total2 = drug2_data.get('HBondDonorCount', 0) + drug2_data.get('HBondAcceptorCount', 0)
+    if abs(hbond_total1 - hbond_total2) > 8:
+        descriptions.append("Differing hydrogen bonding capacity may affect protein binding")
+    
+    # Default description if no specific features stand out
+    if not descriptions:
+        descriptions.append("Potential pharmacokinetic interaction requiring clinical evaluation")
+    
+    return " ".join(descriptions)
+
+def check_drugbank_interaction(drug1_name, drug2_name):
+    """Placeholder function to check DrugBank interactions (you'd need API access)"""
+    # This is a mock function - you'd need actual DrugBank API access
+    drug1_clean = drug1_name.lower().strip()
+    drug2_clean = drug2_name.lower().strip()
+    
+    # Mock known interactions for demonstration
+    known_interactions = {
+        ('warfarin', 'aspirin'): 'Severe: Increased bleeding risk',
+        ('warfarin', 'ibuprofen'): 'Moderate: Increased bleeding risk',
+        ('simvastatin', 'clarithromycin'): 'Severe: Increased risk of myopathy',
+        ('digoxin', 'quinine'): 'Moderate: Increased digoxin levels',
+    }
+    
+    # Check both orders
+    interaction = known_interactions.get((drug1_clean, drug2_clean)) or known_interactions.get((drug2_clean, drug1_clean))
+    
+    if interaction:
+        return interaction
+    else:
+        return "No known interaction in mock database"
+
+def predict_ddi(drug1_name, drug2_name):
+    """Main prediction function"""
+    try:
+        # Fetch data for both drugs
+        drug1_data, error1 = fetch_pubchem_data(drug1_name)
+        drug2_data, error2 = fetch_pubchem_data(drug2_name)
+        
+        if error1 or error2:
+            return f"Error: {error1 or error2}", "", "", "", ""
+        
+        # Generate interaction description
+        interaction_description = generate_interaction_description(drug1_data, drug2_data)
+        
+        # Make prediction
+        result = predictor.predict(interaction_description)
+        
+        # Check DrugBank (mock)
+        drugbank_result = check_drugbank_interaction(drug1_name, drug2_name)
+        
+        # Prepare detailed output
+        drug1_info = f"""
+        **{drug1_name} Properties:**
+        - SMILES: {drug1_data['CanonicalSMILES']}
+        - Molecular Weight: {drug1_data['MolecularWeight']:.2f}
+        - LogP: {drug1_data.get('XLogP', 'N/A')}
+        - TPSA: {drug1_data.get('TPSA', 'N/A')}
+        - H-Bond Donors: {drug1_data.get('HBondDonorCount', 'N/A')}
+        - H-Bond Acceptors: {drug1_data.get('HBondAcceptorCount', 'N/A')}
+        """
+        
+        drug2_info = f"""
+        **{drug2_name} Properties:**
+        - SMILES: {drug2_data['CanonicalSMILES']}
+        - Molecular Weight: {drug2_data['MolecularWeight']:.2f}
+        - LogP: {drug2_data.get('XLogP', 'N/A')}
+        - TPSA: {drug2_data.get('TPSA', 'N/A')}
+        - H-Bond Donors: {drug2_data.get('HBondDonorCount', 'N/A')}
+        - H-Bond Acceptors: {drug2_data.get('HBondAcceptorCount', 'N/A')}
+        """
+        
+        prediction_output = f"""
+        **Generated Interaction Description:**
+        {interaction_description}
+        
+        **AI Prediction:**
+        - Severity: **{result['prediction']}**
+        - Confidence: {result['confidence']:.2%}
+        
+        **Probabilities:**
+        {', '.join([f'{k}: {v:.2%}' for k, v in result['probabilities'].items()])}
+        """
+        
+        return prediction_output, drug1_info, drug2_info, drugbank_result, interaction_description
+        
+    except Exception as e:
+        return f"Error: {str(e)}", "", "", "", ""
+
+# Create Gradio interface
+with gr.Blocks(title="Drug Interaction Severity Predictor", theme=gr.themes.Soft()) as demo:
+    gr.Markdown("# 🧪 Drug Interaction Severity Predictor")
+    gr.Markdown("Predict potential drug-drug interaction severity using molecular properties and AI")
+    
+    with gr.Row():
+        with gr.Column():
+            drug1 = gr.Textbox(label="First Drug Name", placeholder="e.g., Warfarin, Aspirin, Simvastatin...")
+            drug2 = gr.Textbox(label="Second Drug Name", placeholder="e.g., Ibuprofen, Clarithromycin...")
+            predict_btn = gr.Button("Predict Interaction", variant="primary")
+    
+    with gr.Row():
+        with gr.Column():
+            gr.Markdown("## 📊 Prediction Results")
+            prediction_output = gr.Markdown(label="AI Prediction")
+            
+        with gr.Column():
+            gr.Markdown("## 🔍 Drug Properties")
+            with gr.Tab("Drug 1"):
+                drug1_info = gr.Markdown()
+            with gr.Tab("Drug 2"):
+                drug2_info = gr.Markdown()
+    
+    with gr.Row():
+        with gr.Column():
+            gr.Markdown("## 🏥 DrugBank Comparison")
+            drugbank_result = gr.Textbox(label="Known Interaction (Mock Data)", interactive=False)
+        
+        with gr.Column():
+            gr.Markdown("## 📝 Generated Description")
+            interaction_description = gr.Textbox(label="AI-Generated Interaction Description", interactive=False)
+    
+    # Examples
+    gr.Markdown("### 💡 Example Drug Pairs")
+    gr.Examples(
+        examples=[
+            ["Warfarin", "Aspirin"],
+            ["Simvastatin", "Clarithromycin"],
+            ["Digoxin", "Quinine"],
+            ["Metformin", "Ibuprofen"]
+        ],
+        inputs=[drug1, drug2],
+        outputs=[prediction_output, drug1_info, drug2_info, drugbank_result, interaction_description],
+        fn=predict_ddi,
+        cache_examples=True
+    )
+    
+    predict_btn.click(
+        fn=predict_ddi,
+        inputs=[drug1, drug2],
+        outputs=[prediction_output, drug1_info, drug2_info, drugbank_result, interaction_description]
+    )
+
+if __name__ == "__main__":
+    demo.launch(share=True)