Supervised Learning for Gene Expression Analysis

This repository showcases the application of Logistic Regression and Random Forest for gene expression analysis. It automates data processing, model training, and evaluation.

🚀 Features

• Preprocessing: Formats gene expression data
• Machine Learning: Uses Logistic Regression & Random Forest
• Evaluation: Generates accuracy reports and confusion matrices
• Automation: Includes a shell script & GitHub Actions

📂 Repository Structure

📂 **ml_gene_expression_project**  
 ┣ 📂 **data/**               → Stores gene expression data  
 ┣ 📂 **results/**            → Model reports and visualizations  
 ┣ 📜 **preprocess.py**       → Data processing script  
 ┣ 📜 **train.py**            → Model training script  
 ┣ 📜 **evaluate.py**         → Model evaluation with visualization  
 ┣ 📜 **run_pipeline.sh**     → Shell script to automate pipeline execution  
 ┣ 📜 **run_pipeline.yml**    → GitHub Actions workflow  
 ┣ 📜 **app.py**              → Streamlit app for interactive visualization  
 ┣ 📜 **README.md**           → Project documentation  

🏃 Run the Pipeline

bash run_pipeline.sh

🖼️ Sample Outputs

• Model Accuracy Reports in results/
• Confusion Matrices saved as images

🤖 GitHub Actions

This repository supports automatic execution when new data is pushed.

📌 How to Use

1. Clone the repository:

   git clone https://github.com/sivkri/GeneExpression-MachineLearning.git

2. Navigate to the directory:

   cd GeneExpression-MachineLearning

3. Run the pipeline:

   bash run_pipeline.sh

To launch the Streamlit app for interactive visualization:

streamlit run app.py

🔥 Key Enhancements:

✔ Streamlit app is properly emphasized
✔ Instructions for running the app are added
✔ Clear explanation of app features

This version sells your ML + Streamlit project effectively. Let me know if you need further refinements! 🚀

Results & Findings

• Identified top genes differentiating wild-type (WT) vs knockout (KO) conditions
• Evaluated the impact of Eltrombopag (E20) treatment
• Achieved 75% accuracy with the Random Forest classifier

Project Overview

This project applies machine learning techniques to analyze gene expression data under different experimental conditions. Using logistic regression and random forest classifiers, we identify genes differentially expressed due to HuR knockout (ELAVL1 deletion) and Eltrombopag (E20) drug treatment.

Additionally, a Streamlit web application is integrated to provide an interactive visualization of the results.

Citation

If you use this dataset or findings, please cite the following study:
📖 DOI: 10.1186/s12915-025-02131-z

---

Experimental Design

This study investigates how HuR knockout (KO) and Eltrombopag (E20) treatment influence gene expression compared to wild-type (WT) and mock treatment (DMSO).

Sample Groups

The dataset consists of the following experimental conditions:

Sample Group	Description
WT-DMSO	Wild-type (WT) cells treated with mock (DMSO)
WT-E20	Wild-type (WT) cells treated with Eltrombopag (E20)
KO-DMSO	HuR knockout (KO) cells treated with mock (DMSO)
KO-E20	HuR knockout (KO) cells treated with Eltrombopag (E20)

Each sample contains gene expression data across thousands of genes. HuR (ELAVL1) is a key RNA-binding protein, and its knockout may significantly alter gene expression. Eltrombopag is a thrombopoietin receptor agonist that may influence transcriptional programs.

---

Comparisons & Research Questions

I have performed three key comparisons using supervised learning to classify gene expression profiles.

1️⃣ Effect of HuR Knockout (KO vs. WT)

• Comparison: WT-DMSO vs. KO-DMSO
• Objective: Identify genes affected by HuR deletion.
• Machine Learning Approach:
  • Features: Gene expression levels
  • Labels: WT-DMSO (class 0) vs. KO-DMSO (class 1)

2️⃣ Effect of Eltrombopag in Wild-Type Cells

• Comparison: WT-DMSO vs. WT-E20
• Objective: Determine gene expression changes due to Eltrombopag in normal cells.
• Machine Learning Approach:
  • Features: Gene expression levels
  • Labels: WT-DMSO (class 0) vs. WT-E20 (class 1)

3️⃣ Effect of Eltrombopag in HuR Knockout Cells

• Comparison: KO-DMSO vs. KO-E20
• Objective: Understand the HuR-dependent response to Eltrombopag.
• Machine Learning Approach:
  • Features: Gene expression levels
  • Labels: KO-DMSO (class 0) vs. KO-E20 (class 1)

---

Data Processing & Machine Learning Workflow

1. Preprocessing:

   • Normalize expression data
   • Convert into a machine-learning-ready format

2. Model Training & Feature Selection:

   • Train logistic regression and random forest classifiers
   • Perform Principal Component Analysis (PCA)

3. Evaluation:

   • Compute accuracy, confusion matrices, classification reports
   • Identify top differentially expressed genes

4. Visualization & Reporting:

   • Generate PCA scatter plots
   • Save model performance metrics

---

📧 Contact

For queries, feel free to reach out! 🚀

---