decideTestsPerLabel: Decide tests for each label

In scran: Methods for Single-Cell RNA-Seq Data Analysis

Description

Decide which tests (i.e., genes) are significant for differential expression between conditions in each label, using the output of pseudoBulkDGE. This mimics the decideTests functionality from limma.

Usage

decideTestsPerLabel(
  results,
  method = c("separate", "global"),
  threshold = 0.05,
  pval.field = NULL,
  lfc.field = "logFC"
)

summarizeTestsPerLabel(results, ...)

Arguments

results	A List containing the output of pseudoBulkDGE. Each entry should be a DataFrame with the same number and order of rows, containing at least a numeric "PValue" column (and usually a "logFC" column). For summarizeTestsPerLabel, this may also be a matrix produced by decideTestsPerLabel.
method	String specifying whether the Benjamini-Hochberg correction should be applied across all clustesr or separately within each label.
threshold	Numeric scalar specifying the FDR threshold to consider genes as significant.
pval.field	String containing the name of the column containing the p-value in each entry of results. Defaults to "PValue", "P.Value" or "p.value" based on fields in the first entry of results.
lfc.field	String containing the name of the column containing the log-fold change. Ignored if the column is not available Defaults to "logFC" if this field is available.
...	Further arguments to pass to decideTestsPerLabel if results is a List.

Details

If a log-fold change field is available and specified in lfc.field, values of 1, -1 and 0 indicate that the gene is significantly upregulated, downregulated or not significant, respectively. Note, the interpretation of “up” and “down” depends on the design and contrast in pseudoBulkDGE.

Otherwise, if no log-fold change is available or if lfc.field=NULL, values of 1 or 0 indicate that a gene is significantly DE or not, respectively.

NA values indicate either that the relevant gene was low-abundance for a particular label and filtered out, or that the DE comparison for that label was not possible (e.g., no residual d.f.).

Value

For decideTestsPerLabel, an integer matrix indicating whether each gene (row) is significantly DE between conditions for each label (column).

For summarizeTestsPerLabel, an integer matrix containing the number of genes of each DE status (column) in each label (row).

Author(s)

Aaron Lun

See Also

pseudoBulkDGE, which generates the input to this function.

decideTests, which inspired this function.

Examples

example(pseudoBulkDGE)
head(decideTestsPerLabel(out))
summarizeTestsPerLabel(out)

scran documentation built on April 17, 2021, 6:09 p.m.