DRAGON

DRAGON performs joint network analysis of paired methylomic and transcriptomic data using Gaussian Graphical Models (GGMs) to infer regulatory associations and characterize cellular phenotypes and responses to perturbations.

Key Features:

• Paired omics analysis: Analyzes methylomic and transcriptomic profiles collected from the same specimen or cell to model cross-layer interactions.
• Gaussian Graphical Models (GGMs): Employs GGMs to model conditional dependencies and infer network edges across omic layers.
• Parameter calibration: Calibrates parameters to balance network complexity and estimation accuracy while explicitly accounting for characteristics of each omics layer.
• Adaptability and edge recovery: Adapts to variations in edge density and feature size across omic layers, enhancing model inference and edge recovery relative to current state-of-the-art methods.

Scientific Applications:

• Molecular mechanism identification: Analyzes joint transcriptome-methylome data to uncover gene regulation mechanisms such as promoter methylation effects on transcription.
• Biomarker discovery: Has been applied to TCGA breast cancer specimens, where it identified Transcription Factor AP-2 Beta (TFAP2B) as a potential biomarker for basal-type breast cancer.
• Phenotype and perturbation response analysis: Infers regulatory associations that help characterize cellular phenotypes and responses to perturbations.

Methodology:

Integration and joint analysis of multiple omic datasets using Gaussian Graphical Models (GGMs), with parameter calibration to balance network complexity and estimation accuracy and adaptation to edge density and feature size to improve edge recovery.