🧬 Liver Cancer Gene Expression Classification

This repository contains a complete machine learning pipeline for classifying liver cancer based on gene expression data. It includes data preprocessing, exploratory analysis, dimensionality reduction using PCA, and classification using various ML models.

📁 Dataset

The dataset used in this project is assumed to be in CSV format and named liver_cancer_expression_data.csv, containing gene expression features and a type column indicating class labels (e.g., HCC vs. normal).

---

🛠️ Features of the Code

• Missing value handling with column-wise mean imputation.
• Feature normalization using z-score standardization.
• Dimensionality reduction using Principal Component Analysis (PCA).
• Visualizations for data distribution and PCA projection.
• ML models: Random Forest, SVM, XGBoost, Neural Network (MLP).
• Hyperparameter tuning using GridSearchCV.
• Evaluation using accuracy, classification report, confusion matrix, and AUC-ROC.

---

🔧 Requirements

Install the following Python packages before running the script:

pip install pandas numpy matplotlib seaborn scikit-learn xgboost

🧪 How to Run

Place your dataset file as liver_cancer_expression_data.csv in the project directory.

Run the script using:

python liver_cancer_classification.py

---

🧹 Data Preprocessing

• Loaded dataset and checked for null values.
• Imputed missing values with mean (for numeric columns).
• Standardized features using StandardScaler.
• Visualized 20 randomly selected standardized features using boxplots.

---

📊 PCA Visualization

• Reduced data to 2D using PCA for visualization.
• Plotted principal components with color-coded classes.

---

⚙️ Machine Learning Models

Trained and evaluated the following models:

Model	Hyperparameters Tuned	Evaluation Metrics
Random Forest	n_estimators, max_depth, min_samples_split	Accuracy, AUC-ROC
SVM	C, kernel	Accuracy, AUC-ROC
XGBoost	n_estimators, max_depth	Accuracy, AUC-ROC
Neural Network	alpha (regularization)	Accuracy, AUC-ROC

Note: Dimensionality reduction to 50 principal components was performed before training to avoid overfitting.

---

📈 Model Evaluation

• Accuracy and classification reports printed for each model.
• ROC curves plotted with AUC scores for probabilistic models.

---

📌 Example Output (Console)

Random Forest Test Accuracy: 0.8750

📊 Classification Report

Class	Precision	Recall	F1-score	Support
Normal	0.89	0.86	0.87	14
HCC	0.86	0.89	0.88	14

AUC-ROC Score for Random Forest: 0.93

📎 Notes

• Set random seed for reproducibility.
• Suitable for high-dimensional datasets with limited samples.
• Uses stratify=y in train-test split to preserve class balance.

---

🧠 Author

Divya Karade
Chemoinformatics & ML in Drug Discovery
LinkedIn | GitHub

---

📄 License

This project is open-source and available under the MIT License.