omicwas

omicwas performs cell-type-specific association analysis on bulk omics data by inferring cell composition and applying nonlinear regression with ridge regularization to detect differential DNA methylation and gene expression associated with traits.

Key Features:

• Cell-Type-Specific Analysis: Infers cell type composition from bulk omics data and estimates cell-type-specific effects on omic measures.
• Nonlinear Regression Approach: Models omic measures on logit/log scales while treating cell composition on a linear scale using nonlinear regression to analyze effects across different scales.
• Ridge Regularization to Address Multicollinearity: Applies ridge regularization to interaction terms between cell type proportions and traits to mitigate multicollinearity and stabilize estimates.
• Balanced Performance Metrics: Nonlinear ridge regression demonstrates balanced sensitivity, specificity, and precision in simulations, achieving higher precision than marginal models and enabling detection of weak signals in real data.
• Implementation: Implemented in R.

Scientific Applications:

• Epigenome-Wide Association Studies (EWAS): Facilitates identification of epigenetic modifications associated with specific traits across cell types within bulk tissue samples.
• Differential Gene Expression Analysis: Discriminates gene expression changes attributable to particular cell types in bulk samples.
• Quantitative Trait Loci (QTL) Analyses: Integrates cell-type-specific information with QTL data to examine genetic associations in the cellular context.

Methodology:

Infers cell type proportions from bulk omics data and applies nonlinear regression with ridge regularization to model interactions between inferred cell type compositions and traits, accounting for logit/log versus linear scales and multicollinearity.