discordant

discordant identifies differential correlation patterns among molecular feature pairs in -omics datasets by categorizing per-group correlation types to detect cases where pairs are correlated in one phenotype but not another.

Key Features:

• Differential correlation analysis: Identifies molecular feature pairs with distinct correlation patterns between groups, including cases of correlation in one group but not the other.
• Per-group correlation categorization: Categorizes the types of correlations present within each dataset group separately.
• Mixture models: Employs mixture models to model correlation category assignments.
• Comparative evaluation: Compares performance with existing differential correlation methods using simulations and application to two diverse -omics datasets.
• Phenotype-related feature detection: Demonstrates improved identification of phenotype-related features at rates comparable or superior to other methods.
• Computational efficiency: Performs analyses while maintaining computational efficiency.

Scientific Applications:

• Biomarker discovery: Aids in uncovering potential biomarkers from differential correlation patterns in -omics datasets.
• Mechanistic inference: Supports understanding underlying molecular interactions and mechanisms that vary between phenotypes.
• Cross-phenotype interaction analysis: Detects molecular feature pairs exhibiting condition-specific correlations across biological conditions.

Methodology:

Models per-group correlation categories using mixture models and evaluates performance via simulations and application to two diverse -omics datasets while comparing to existing differential correlation methods.