DeweyWang/ETH_ML4G_Project-1

Gene Expression Prediction Dataset

📌 Overview

This dataset is designed for predicting gene expression levels from chromatin landscape data, including histone modifications and chromatin accessibility.

It is part of a machine learning project in genomics, where the goal is to model the relationship between epigenetic signals and gene expression.

👉 Full project code (including preprocessing and prediction): https://github.com/Dewey-Wang/Gene-expression-prediction/tree/main

---

📂 Dataset Structure

The dataset consists of two main components:

1. Raw Data

• Total size: 18.66 GB
• Number of files: 72
• Includes:
  • Histone modification data (ChIP-seq)
  • Chromatin accessibility (DNase-seq)
  • Gene expression (CAGE)
  • Gene annotation (TSS, gene body, RefSeq)

---

2. Preprocessed Data

• Total size: 6.36 GB
• Number of files: 53
• Includes:
  • Feature matrices for machine learning
  • Aggregated signals around genomic regions (e.g. TSS windows)
  • Normalized inputs ready for model training

👉 Full preprocessing code is available in the GitHub repository above.

---

🎯 Task

The main task is:

Predict gene expression levels from chromatin features

• Input: epigenetic signals (ChIP-seq, DNase-seq)
• Output: gene expression values

---

📊 Evaluation

Typical evaluation metrics:

• Spearman correlation (primary)
• Pearson correlation
• R² score

---

🧬 Data Details

• Genome version: hg38 / GRCh38
• Multiple cell lines included
• Data normalized for cross-cell-line comparison

---

🚀 Usage

You can either:

1. Use preprocessed data directly for ML models
2. Reproduce preprocessing using provided code

---

⚠️ Notes

• Raw data is large (~18.66 GB)
• Preprocessed data is recommended for quick experimentation
• Suitable for machine learning and bioinformatics research