runPCA: Run PCA scores.

In nlhuong/TimeSeriesExperiment: Analysis for short time-series data

Description

Compute the projection of either samples or time series features onto a PCA space. In case, of PCA for features, the PCA can be computed for individual sample group as indicated or for samples from all groups. In either, case the data is first collapsed over replicates, so that each gene is represented as a vector of a single time course.

Usage

runPCA(
  object,
  collapse.replicates = FALSE,
  groups.selected = NULL,
  var.stabilize.method = "log1p"
)

Arguments

object	A TimeSeriesExperiment object
collapse.replicates	Whether PCA should be computed on the data with replicates aggregated.
groups.selected	An optional character string indicating a particular group of samples PCA should be applied to. By default set to NULL and all groups are included.
var.stabilize.method	Method for variance stabilization (VST). Currently, supports "none" (no VST), "log1p" (log plus one), "asinh" (inverse hyperbolic sine) or "deseq" (varianceStabilizingTransformation function from DESeq2 package). Default is "log1p".

Value

Returns TimeSeriesExperiment object with PCA results in dim.red slot, a lists containing matrices of coordinates 'pca_sample', and 'pca_features', as well as a vector 'pca_eigs'.

Examples

data("endoderm_small")
endoderm_small <- runPCA(endoderm_small)
head(dimensionReduction(endoderm_small, "pca_sample")[, 1:5])
head(dimensionReduction(endoderm_small, "pca_eigs"))

nlhuong/TimeSeriesExperiment documentation built on Sept. 8, 2021, 12:30 a.m.