Skip to contents

Generate Functional Spectra

Source: R/FunctionalSpectra.R
getFunctionalSpectra.Rd

This function generates functional spectra for given gene expression profiles. Uses a fast batch native kernel by default, with automatic fallback to the legacy per-row parallel path when the input structure prevents batch indexing.

Usage

getFunctionalSpectra(eps, geneSets = "MSigDB", scale = TRUE, cores = NULL)

Arguments

eps

a data.frame containing gene expression profiles (each row presents one sample)

geneSets

gene sets to score. Either `"MSigDB"` for the bundled cache, `"MSigDB_<version>"` for an explicit bundled release, or a named list such as the result of [get_msigdbr()]

scale

logical indicating whether to center each gene column (default: TRUE)

cores

integer or NULL; number of CPU cores. NULL uses the native kernel's thread pool. Set to a specific integer for legacy parallel fallback.

Value

a data.frame containing functional spectra

Details

A functional spectrum summarises, for one sample, how strongly each gene set is enriched among that sample's most highly expressed genes.

With `scale = TRUE` (the default) each gene is centred across the samples in `eps` by subtracting its mean, so every value becomes a deviation from the cohort average for that gene. Each sample is then ranked independently on those centred values and scored with the weighted running-sum enrichment statistic.

This is not a log fold change. DeepCC does not compare labelled groups, does not select differentially expressed genes, and needs no group labels. A score is relative to the other samples supplied in `eps`, so the same sample receives different scores in a different cohort. Supply the full cohort you wish to compare against, and keep that cohort fixed between training and prediction.

With `scale = FALSE` no centring is applied and each sample is ranked on its own values, which makes a score independent of the other rows.

A single row with `scale = TRUE` centres to zero and therefore returns all zeros. Use [getFunctionalSpectrum()] for genuine single-sample scoring.

See also

getFunctionalSpectrum for a single expression profile.

Examples

if (FALSE) { # \dontrun{
set.seed(42)
eps <- as.data.frame(matrix(rnorm(10*100), nrow=10, ncol=100))
colnames(eps) <- paste0("G", seq_len(100))
fs <- getFunctionalSpectra(eps, geneSets=list(setA=c("G1","G5","G20")))
} # }