Bayesian Epiallele Detection (BED)

Bayesian Epiallele Detection (BED) infers and characterizes tumor epialleles from reduced representation bisulfite sequencing (RRBS) across multiple tumor regions to quantify epigenetic heterogeneity and support studies of tumor evolution and biomarker discovery.

Key Features:

• Bayesian Inference Model: Employs a Bayesian framework that integrates data across multiple tumor regions to improve inference of epiallele presence and distribution.
• Automatic Parameter Estimation: Automatically infers the total number of epialleles, their DNA methylation patterns, and a noise hyperparameter directly from RRBS data.
• Incorporation of Uncertainty: Models uncertainty about the origin of sequencing reads by marginalizing over posterior densities.
• Contamination Correction: Estimates and corrects for contamination from normal tissue within tumor samples.
• Phylogenetic Analysis: Traces epiallele distributions across regions to facilitate reconstruction of tumor evolutionary relationships.
• Identification of Differentially Expressed Epialleles: Detects epialleles that differ between matched normal and cancerous tissues.
• Measure of Global Epigenetic Disorder: Provides a quantitative measure of global epigenetic heterogeneity within tumors.

Scientific Applications:

• Tumor Evolution Studies: Analysis of epiallele distributions to infer intratumoral evolutionary trajectories and treatment resistance mechanisms.
• Biomarker Discovery: Identification of epialleles that are differentially present between normal and cancerous tissue for potential diagnostic or therapeutic markers.
• Epigenetic Research: Detailed characterization of DNA methylation haplotypes (epialleles) to study epigenetic dynamics beyond single-site DNAm measures.

Methodology:

Processes RRBS data from multiple tumor regions and a matched normal sample and applies Bayesian inference with marginalization over posterior densities to infer epialleles while automatically estimating epiallele number, methylation patterns, a noise hyperparameter, and correcting for normal contamination.