markerEnrichment: Find enriched markers per identified cluster and calculate...
In kevinblighe/scToolkit: scDataviz: single cell dataviz and downstream analyses

Description Usage Arguments Details Value Author(s) Examples

Find enriched markers per identified cluster and calculate cluster abundances across these for samples and metadata variables.

markerEnrichment(
  indata,
  meta = NULL,
  assay = "scaled",
  sampleAbundances = TRUE,
  sampleID = "sample",
  studyvarID = NULL,
  clusterAssign = metadata(indata)[["Cluster"]],
  funcSummarise = function(x) mean(x, na.rm = TRUE),
  method = "Z",
  prob = 0.1,
  limits = c(-1.96, 1.96),
  verbose = TRUE
)

`indata`	A data-frame or matrix, or `SingleCellExperiment` object. If a data-frame or matrix, this should relate to expression data (cells as columns; genes as rows). If a `SingleCellExperiment` object, data will be extracted from an assay component named by `assay`.
`meta`	If 'indata' is a non-`SingleCellExperiment` object, `meta` must be activated and relate to a data-frame of metadata that aligns with the columns of `indata`, and that also contains a column name specified by `studyvarID`.
`assay`	Name of the assay slot in `indata` from which data will be taken, assuming `indata` is a `SingleCellExperiment` object.
`sampleAbundances`	Logical, indicating whether or not to calculate cluster abundances across study samples.
`sampleID`	If `sampleAbundances == TRUE`, a column name from the provided metadata representing over which sample cluster abundances will be calculated.
`studyvarID`	A column name from the provided metadata representing a condition or trait over which cluster abundances will be calculated.
`clusterAssign`	A vector of cell-to-cluster assignments. This can be from any source but must align with your cells / variables. There is no check to ensure this when 'indata' is not a `SingleCellExperiment` object.
`funcSummarise`	A mathematical function used to summarise expression per marker per cluster.
`method`	Type of summarisation to apply to the data for final marker selection. Possible values include `Z` or `quantile`. If `Z`, `limits` relate to lower and upper Z-score cut-offs for low\|high markers. The defaults of -1.96 and +1.96 are equivalents of p<0.05 on a two-tailed distribution. If `quantile`, `prob` will be used to define the `n`th lower and 1 - `n`th upper quantiles, which will be used for selecting low\|high markers.
`prob`	See details for `method`.
`limits`	See details for `method`.
`verbose`	Boolean (TRUE / FALSE) to print messages to console or not.

Find enriched markers per identified cluster and calculate cluster abundances across these for samples and metadata variables. markerEnrichment first collapses your input data's expression profiles from the level of cells to the level of clusters based on a mathematical function specified by funcSummarise. It then either selects, per cluster, low|high markers via quantiles, or transforms this collapsed data to global Z-scores and selects low|high markers based on Z-score cut-offs.

A data.frame object.

Kevin Blighe <kevin@clinicalbioinformatics.co.uk>

# create random data that follows a negative binomial
mat <- jitter(matrix(
  MASS::rnegbin(rexp(1000, rate=.1), theta = 4.5),
  ncol = 20))
colnames(mat) <- paste0('CD', 1:ncol(mat))
rownames(mat) <- paste0('cell', 1:nrow(mat))

u <- umap::umap(mat)$layout
colnames(u) <- c('UMAP1','UMAP2')
rownames(u) <- rownames(mat)
clus <- clusKNN(u)

metadata <- data.frame(
  group = c(rep('PB1', 25), rep('PB2', 25)),
  row.names = rownames(u))

markerEnrichment(t(mat), meta = metadata,
  sampleAbundances = FALSE,
  studyvarID = 'group', clusterAssign = clus)