normalizeVariance: Identify overdispersed genes by normalizing counts per...

In JEFworks-Lab/veloviz: VeloViz: RNA-velocity informed 2D embeddings for visualizing cell state trajectories

Identify overdispersed genes by normalizing counts per million (CPM) gene expression variance relative to transcriptome-wide expectations (Modified from SCDE/PAGODA2 code)

Description

Normalizes gene expression magnitudes to with respect to its ratio to the transcriptome-wide expectation as determined by local regression on expression magnitude

Usage

normalizeVariance(
  cpm,
  gam.k = 5,
  alpha = 0.05,
  max.adjusted.variance = 1000,
  min.adjusted.variance = 0.001,
  verbose = TRUE,
  plot = FALSE,
  details = FALSE
)

Arguments

cpm	Counts per million (CPM) matrix. Rows are genes, columns are cells.
gam.k	Generalized additive model parameter; the dimension of the basis used to represent the smooth term (default: 5)
alpha	Significance threshold for overdispersed genes (default: 0.05)
max.adjusted.variance	Ceiling on maximum variance after normalization to prevent infinites (default: 1e3)
min.adjusted.variance	Floor on minimum variance after normalization (default: 1e-3)
verbose	Boolean for verbosity setting (default: TRUE)
plot	Boolean to plot mean variance plots before and after correction
details	Boolean to return detailed data frame or names of genes (default: FALSE)

Value

A list with two items: (1) an adjusted CPM matrix with the same dimensions as the input and (2) a dataframe with the summary statistics for each gene.

Examples

data(vel)
curr <- vel$current

normalizeDepth(curr)

JEFworks-Lab/veloviz documentation built on Sept. 14, 2022, 4:03 p.m.