Skip to contents

NormalizeCounts

Source: R/Pseudobulk.R
NormalizeCounts.Rd

Perform per-sample normalization of a count assay stored in a SummarizedExperiment and add the normalized matrix as a new assay.

Usage

NormalizeCounts(
  se,
  method = c("CPM", "logCPM", "logNorm", "VST", "rlog"),
  assay_name = "counts",
  new_assay_name = NULL,
  pseudocount = 1,
  ...
)

Arguments

se

A SummarizedExperiment containing a raw counts assay.

method

Character; one of "CPM", "logCPM", "logNorm", "VST", "rlog".

  • "CPM": counts per million.

  • "logCPM": log2(CPM + pseudocount).

  • "logNorm": Seurat-style log1p(counts / size_factor * 1e4) where size_factor = colSums(counts) / median(colSums(counts)).

  • "VST", "rlog": variance-stabilizing transform or rlog via DESeq2.

assay_name

Character scalar; name of the assay in se to normalize (default: "counts").

new_assay_name

Character or NULL; name to assign the normalized assay. If NULL, defaults to the chosen method.

pseudocount

Numeric scalar added to CPM before log2 in "logCPM" (default: 1).

...

Additional arguments forwarded to DESeq2::vst or DESeq2::rlog when method is "VST" or "rlog".

Value

A SummarizedExperiment identical to se but with a new assay named new_assay_name containing the normalized matrix.

Details

The function checks that se is a SummarizedExperiment and that assay_name exists and is a (possibly sparse) matrix. For "VST" and "rlog", DESeq2 must be installed; the function converts the assay to a dense matrix and constructs a DESeqDataSet with design ~ 1 before applying the transform. Errors are raised for invalid inputs.

See also

DESeq2::vst, DESeq2::rlog

Examples

# Basic usage (assuming `se` is a SummarizedExperiment with a "counts" assay)
# se_norm <- NormalizeSE(se, method = "logCPM")