EAGLE

EAGLE predicts tissue- and cell type-specific enhancer–gene (EG) interactions to infer long-range regulatory relationships between distal enhancers and their target genes using a minimal set of genomic and expression-derived features.

Key Features:

• Minimal feature set: Employs six key features derived from enhancer genomic data and gene expression datasets.
• Tissue/cell-type specificity: Predicts EG interactions across diverse tissue and cell types without requiring histone modification or DNA methylation data.
• Cross-validation performance: Demonstrates improved accuracy over existing methods through rigorous cross-validation.
• Enrichment validation: Enrichment analyses of predicted interacting pairs highlight transcriptional factors, epigenetic modifications, and expression quantitative trait loci (eQTLs) that distinguish interacting from non-interacting pairs.
• Species-scale application: Applied to mouse and human genomes, producing millions of EG interactions (mouse: 7,680,203 interactions involving 31,375 genes and 138,547 enhancers across 89 tissue/cell types; human: 7,437,255 interactions involving 43,724 genes and 177,062 enhancers across 110 tissue/cell types).
• Reported resource: Predicted EG interaction datasets are reported in enhanceratlas.org.

Scientific Applications:

• Regulatory genomics: Maps distal enhancer–target gene links to study mechanisms of long-range gene regulation.
• Functional annotation: Links predicted EG pairs to transcriptional factors, epigenetic modifications, and eQTLs for downstream interpretation of regulatory signals.
• Comparative tissue mapping: Generates large-scale EG interaction maps across 89 mouse and 110 human tissue/cell types for cross-species and tissue-specific analyses.

Methodology:

Uses the EAGLE algorithm (Enhancer And Gene Learning Ensemble) based on six features derived from enhancer genomic data and gene expression datasets, validated by cross-validation and assessed by enrichment analyses for transcription factors, epigenetic modifications, and eQTLs.