Merges similar transcriptional profiles into representative 'pools'

applyMicroClustering(
  exprData,
  cellsPerPartition = 10,
  filterInput = "fano",
  filterThreshold = round(ncol(exprData) * 0.05),
  filterNumMad = 2,
  latentSpace = NULL,
  K = round(sqrt(ncol(exprData)))
)

Arguments

exprData

the expression data matrix

cellsPerPartition

control over the target number of cells to put into each supercell

filterInput

name of filtering method ('threshold' or 'fano') or list of genes to use when computing projections.

filterThreshold

Threshold to apply when using the 'threshold' or 'fano' projection genes filter. If greater than 1, this specifies the number of cells in which a gene must be detected for it to be used when computing PCA. If less than 1, this instead specifies the proportion of cells needed

filterNumMad

Number of median absolute deviations to use when selecting highly-variable genes in each mean-sorted bin of genes

latentSpace

(Optional) Latent space to be used instead of PCA numeric matrix cells x components

K

Number of neighbors to use for finding pools.

Value

pooled cells - named list of vectors - cells in each supercell

Details

A latent space is computed for the expression data via PCA after filtering on genes (using parameters filterInput and filterThreshold).

Alternately, a latent space can be supplied via the latentSpace argument

Euclidean distance within the latent space is then used to create cell pools