deswish.RdThe DESeq2-apeglm With Inferential Samples implementation supposes
a hierarchical distribution of log2 fold changes.
The final posterior standard deviation is calculated by
adding the posterior variance from modeling biological replicates
computed by apeglm, and the observed variance on the posterior mode
over inferential replicates. This function requires the DESeq2 and
apeglm packages to be installed and will print an error if they are
not found.
deswish(y, x, coef)a SummarizedExperiment containing the inferential
replicate matrices, as output by tximeta, and
then with labelKeep applied. One does not need to
run scaleInfReps as scaling is done internally via
DESeq2.
the design matrix
the coefficient to test (see lfcShrink)
a SummarizedExperiment with metadata columns added: the log2 fold change and posterior SD using inferential replicates, and the original log2 fold change (apeglm) and its posterior SD
The DESeq and lfcShrink function in the DESeq2 package:
Zhu, Ibrahim, Love "Heavy-tailed prior distributions for sequence count data: removing the noise and preserving large differences" Bioinformatics (2018).
Love, Huber, Anders "Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2" Genome Biology (2014).
# a small example... 500 genes, 10 inf reps
y <- makeSimSwishData(m=500, numReps=10)
y <- labelKeep(y)
#y <- deswish(y, ~condition, "condition_2_vs_1")