voomCLR allows effective differential cell composition analysis in
cell type count data. It leverages compositional transformations, and
adopts bias correction on the estimated effect sizes to correct for
compositional bias induced by such transformation. The uncertainty
involved in estimating the bias can be propagated in the statistical
inference via bootstrapping. Additionally, it accommodates proper
modeling of the mean-variance structure of the counts.
voomCLR relies on the limma package, and in fact re-uses code chunks
from the limma R package, which is available on Bioconductor at
https://bioconductor.org/packages/release/bioc/html/limma.html.
Install the development version from GitHub using
devtools::install_github("koenvandenberge/voomCLR")The vignette of voomCLR walks you through the basics of using voomCLR.