User-Guided Visual Analytics of Genome-Wide DNA Methylation Data Based on Self-Organizing Maps

Area of Science:

• Epigenetics
• Bioinformatics
• Computational Biology

Background:

• DNA methylation is crucial in disease, especially cancer.
• High-throughput methylation data presents analysis challenges.
• Existing tools lack interactivity and machine learning integration.

Purpose of the Study:

• To develop an interactive framework for epigenomic data exploration.
• To enable interpretable visualization and machine learning on reduced feature spaces.
• To identify disease-associated methylation signatures.

Main Methods:

• Utilized Self-Organizing Maps for data exploration.
• Introduced 'meta sites' for clustering CpG sites.
• Integrated dimensionality reduction (PCA, t-SNE, UMAP) and logistic regression.

Main Results:

• Developed a framework for real-time, interpretable visualization.
• Generated metasite relevance maps highlighting discriminative epigenetic patterns.
• Demonstrated utility in analyzing pheochromocytoma/paraganglioma methylation data.

Conclusions:

• The framework facilitates visually driven discovery of co-regulated modules.
• Offers an intuitive interface for exploring complex methylation landscapes.
• Aids in identifying disease-specific epigenetic biomarkers.