DNA Methylation Data Epigenetic Biomarkers

Dataset Description

About This Dataset

This dataset is designed for researchers, bioinformaticians, and machine learning practitioners working on:

• Epigenetic biomarker discovery
• DNA methylation analysis
• Retinal disease prediction
• Binary classification tasks in genomics
• Feature-based machine learning in bioinformatics

The dataset focuses on CpG (Cytosine-phosphate-Guanine) sites and their methylation patterns, which play crucial roles in gene regulation and disease development. Retinitis Pigmentosa is a group of rare genetic disorders that involve progressive vision loss, and epigenetic factors like DNA methylation may serve as early detection biomarkers.

---

Dataset Statistics

• Total Records: 1,000 samples
• Total Features: 5 columns
• Target Variable: methylation_status (Binary: 0 = Unmethylated, 1 = Methylated)
• Class Distribution:
  • Methylated (1): 554 samples (55.4%)
  • Unmethylated (0): 446 samples (44.6%)
• Missing Values: None (Complete dataset)
• Data Quality: High-quality, normalized, ready for ML/DL applications

---

Column Descriptions

1. cpg_density (float64)

• Description: Density of CpG sites in the genomic region
• Range: 0.052 to 1.000
• Mean: 0.698
• Standard Deviation: 0.183
• Interpretation: Higher values indicate CpG-rich regions (CpG islands), which are often associated with gene promoters

2. genomic_location (float64)

• Description: Normalized genomic position within the chromosome or region
• Range: 0.000 to 1.000
• Mean: 0.612
• Standard Deviation: 0.234
• Interpretation: Represents the relative position of the CpG site; 0 = start, 1 = end of region

3. regulatory_score (float64)

• Description: Score indicating the regulatory potential of the genomic region
• Range: 0.000 to 1.000
• Mean: 0.460
• Standard Deviation: 0.291
• Interpretation: Higher scores suggest regions with higher regulatory activity (e.g., promoters, enhancers)

4. conservation_score (float64)

• Description: Evolutionary conservation score across species
• Range: 0.064 to 1.000
• Mean: 0.643
• Standard Deviation: 0.199
• Interpretation: Higher scores indicate regions conserved across evolution, suggesting functional importance

5. methylation_status (int64) - TARGET VARIABLE

• Description: Binary label indicating DNA methylation status
• Values:
  • 0: Unmethylated (44.6% of samples)
  • 1: Methylated (55.4% of samples)
• Interpretation: Indicates whether the CpG site is methylated, which can affect gene expression

---

Use Cases and Applications

1. Binary Classification

• Predict methylation status based on genomic features
• Suitable for: Logistic Regression, SVM, Random Forest, XGBoost, Neural Networks

2. Feature Importance Analysis

• Identify which genomic features most strongly predict methylation status
• Tools: SHAP, LIME, feature permutation importance

3. Biomarker Discovery

• Identify epigenetic patterns associated with Retinitis Pigmentosa
• Potential for early disease detection and risk assessment

4. Model Benchmarking

• Compare performance of different ML algorithms on genomic data
• Baseline accuracy: ~97% (as demonstrated in the accompanying project)

5. Educational Purposes

• Teach bioinformatics and machine learning concepts
• Demonstrate real-world genomic data analysis workflows

---

Data Format

• File Format: CSV (Comma-Separated Values)
• File Size: ~80 KB
• Encoding: UTF-8
• Delimiter: Comma (,)
• Header: Yes (first row contains column names)
• Ready for: Python (pandas), R, MATLAB, Excel