IMIX

IMIX implements a multivariate mixture model framework for integrative analysis of DNA methylation, copy number variation (CNV), and gene expression to identify coordinated multi-omic signals while maintaining statistical power and controlling the overall false discovery rate (FDR).

Key Features:

• Multivariate Mixture Model Framework: Integrates DNA methylation, CNV, and gene expression simultaneously using a multivariate mixture model to detect coordinated signals across data types.
• Relaxation of Conditional Independence Assumption: Examines and relaxes the conditional independence assumption to allow dependence among data types and better capture biological interactions.
• Statistical Rigor: Implements statistically-principled model selection and overall FDR control and achieves lower misclassification rates in simulations compared with Benjamini-Hochberg FDR control, the q-value, and local FDR.
• Computational Efficiency: Optimized for large-scale genomic datasets and applicable to genome-wide studies such as GWAS.

Scientific Applications:

• TCGA multi-omic analysis: Applied to The Cancer Genome Atlas (TCGA) datasets to integrate methylation, CNV, and expression data.
• Bladder Cancer Subtyping: Facilitated discovery of multi-omic insights into luminal and basal subtypes of bladder cancer.
• Pancreatic Cancer Prognosis: Contributed to identification of prognostic markers for pancreatic cancer.

Methodology:

Applies a multivariate mixture model, relaxes the conditional independence assumption, uses statistically-principled model selection and overall FDR control, and evaluates performance via extensive simulations and comparisons to Benjamini-Hochberg FDR control, the q-value, and local FDR.