restrictCorpus: Restrict to informative words (genes) for topic modeling

In JEFworks-Lab/STdeconvolve: Reference-free Cell-Type Deconvolution of Multi-Cellular Spatially Resolved Transcriptomics Data

Restrict to informative words (genes) for topic modeling

Description

identifies over dispersed genes across pixels to use as informative words (genes) in topic modeling. Also allows ability to restrict over dispersed genes to those that occur in more than and/or less than selected fractions of pixels in corpus. Limits to the top 1000 overdispersed genes in order to keep the corpus to a reasonable size.

Usage

restrictCorpus(
  counts,
  removeAbove = 1,
  removeBelow = 0.05,
  alpha = 0.05,
  nTopOD = 1000,
  plot = FALSE,
  verbose = TRUE
)

Arguments

counts	genes x pixels gene count matrix
removeAbove	remove over dispersed genes that are present in more than this fraction of pixels (default: 1.0)
removeBelow	remove over dispersed genes that are present in less than this fraction of pixels (default: 0.05)
alpha	alpha parameter for getOverdispersedGenes(). Higher = less stringent and more overdispersed genes returned (default: 0.05)
nTopOD	number of top over dispersed genes to use. int (default: 1000). If the number of overdispersed genes is less then this number will use all of them, or set to NA to use all overdispersed genes.
plot	return histogram plots of genes per pixel and pixels per genes for over dispersed genes and after corpus restriction. (default: FALSE)
verbose	(default: TRUE)

Value

a gene by pixel matrix where the remaining genes have been filtered

Examples

data(mOB)
corpus <- restrictCorpus(counts = mOB$counts)
corpus

JEFworks-Lab/STdeconvolve documentation built on April 14, 2023, 1:04 p.m.