Nonparametric analysis of contributions to variance in genomics and epigenomics data

Functional genomics studies, despite increasingly varied assay types and complex experimental designs, are typically analyzed by methods that are unable to identify confounding effects and that incorporate parametric assumptions particular to gene expression data. We present MAVRIC, a nonparametric method to quantify variance explained by experimental covariates and perform differential analysis on arbitrary data types. We demonstrate that MAVRIC can accurately associate covariates with underlying data variance, deliver sensitive and specific identification of genomic loci with differential counts, and provide effective noise reduction of large-scale consortium data sets.