Drug Biomarker Prediction

This project predicts drug response (biomarker values) from molecular data using a machine learning approach based on RDKit molecular descriptors and XGBoost regression. It also provides a user-friendly web interface using Streamlit for interactive predictions.

---

📁 File Structure

├── drug_biomarker_model.py         # Training script for XGBoost model
├── valid_data_prediction.py        # Prediction script using trained model
├── xgb_trained_model.json          # Trained XGBoost model
├── app.py                         # Streamlit web app for interactive prediction
├── submission.csv                  # Final output with predicted biomarker values
├── train.csv                       # Training dataset (required by training script)
├── valid.csv                       # Validation dataset (required by prediction script)
├── requirements.txt                # Python dependencies

---

📦 Requirements

Install the required packages:

pip install -r requirements.txt

---

🧠 Model Training (drug_biomarker_model.py)

This script trains an XGBoost regression model using the train.csv dataset. It:

• Extracts Morgan fingerprints and molecular descriptors from SMILES.
• Concatenates them as features.
• Trains an XGBRegressor.
• Evaluates the model using MAE and R² on train/test sets.
• Saves the model as xgb_trained_model.json.

Run with:

python drug_biomarker_model.py

---

🔍 Drug Response Prediction (valid_data_prediction.py)

This script predicts drug response values using a trained model. It:

• Loads the trained model from xgb_trained_model.json.
• Reads a validation CSV (valid.csv) with columns like Drug_ID and Drug.
• Computes Morgan fingerprints and molecular descriptors.
• Generates predictions for the Bio_Marker_Value.
• Saves the output to predicted_biomarker_values.csv.

Run with:

python valid_data_prediction.py

---

🌐 Streamlit Web App (app.py)

The Streamlit app provides an interactive web interface for predicting biomarker values from SMILES strings.

How to Run

1. Make sure all dependencies are installed (see Requirements).

2. Run the following command in your project directory:

   streamlit run app.py

3. A browser window will open. Enter a SMILES string to get the predicted bio marker value.

---

📄 Output

The final predictions are saved in:

submission.csv

It contains the columns:

• Drug_ID
• Drug
• Bio_Marker_Value (predicted)

This file can be directly used as the assignment submission.

---

📌 Notes

• Ensure train.csv and valid.csv are present in the same directory before running the scripts.
• The descriptor set used includes:
  • Molecular Weight (MolWt)
  • LogP
  • Topological Polar Surface Area (TPSA)
  • Number of H-Bond Donors (HBD)
  • Number of H-Bond Acceptors (HBA)
• The Streamlit app requires the trained model file (xgb_trained_model.json) to be present in the project directory.