plotMarkerHeat: plotMarkerHeat
In mahmoudibrahim/genesorteR: Feature Ranking in Clustered Single Cell Data

Description Usage Arguments Details Value Author(s) See Also Examples

View source: R/plotMarkerHeat.R

Plot a heatmap of expression values for a select set of genes across all cells. Optionally also cluster those genes.

plotMarkerHeat(
  exp,
  classes,
  markers,
  colors = colorRampPalette(rev(c("orangered4", "orangered", "gray90", "dodgerblue",
    "dodgerblue4")))(n = 100),
  newOrder = 1:length(unique(classes)),
  clusterGenes = FALSE,
  clusterGenesK = length(unique(classes)),
  averageCells = 0,
  outs = FALSE,
  plotheat = TRUE,
  gaps = TRUE,
  seed = 10
)

`exp`	A matrix of expression values. Typically the one supplied to `sortGenes`.
`classes`	A vector or factor of cell classes, whose length is equal to `ncol(exp)`.
`markers`	A character vector of gene names to plot in the heatmap.
`colors`	Color palette used for the heatmap.
`newOrder`	Reorder the clusters in the heatmap? See Examples.
`clusterGenes`	Cluster genes before plotting?
`clusterGenesK`	How many clusters should genes by clustered into? See Details.
`averageCells`	Plot averages of cells instead of individual cells. You can use this when you have a large number of cells. See Details.
`outs`	Should gene cluster output and pheatmap object be returned? FALSE by default.
`plotheat`	Should the heatmap be drawn? TRUE by default.
`gaps`	Should the heatmap have gaps between cell types and gene clusters? TRUE by default.
`seed`	Randomization seed used for gene clustering initialization.

clusterGenesK parameter controls the number of gene clusters when clusterGenes is TRUE. By default, it is equal to the number of cell types/clusters. This is usually a good initial guess, the optimal number of clusters (what does that even mean?) is typically somewhere around this value in my subjective experience, unless there are many cell clusters.

By default, the heatmap plots every single cell in one column, this might take forever if you have a lot of cells (would say >10k) or it can crash when you do not have enough RAM. If so, it might be good to set averageCells to n where n is the number of cells you want to average. For example, if averageCells = 10, every 10 cells will be averaged (without averaging across cell clusters) before plotting the heatmap. if averageCells =< 1, no averaging happens. Hints: (1) If you want one column per cell cluster, set averageCells to a very high number (larger than the number of cells in the largest cluster). (2) Gene clustering occurs after cell averaging, so averaging might be useful if the gene clusters you get do not make much sense without averaging (averaging will smooth out zeros).

If outs is TRUE, plotMarkerHeat returns a list containing:

`p`	The pheatmap object corresponding to the plot.
`gene_class_info`	Gene cluster assignments if gene clustering was requested. See Details.
`new_class_info`	The new order of cell clusters. See Examples.

Mahmoud M Ibrahim <mmibrahim@pm.me>

plotTopMarkerHeat

data(kidneyTabulaMuris)
gs = sortGenes(kidneyTabulaMuris$exp, kidneyTabulaMuris$cellType)
mm = getMarkers(gs, quant = 0.999)

#this plots a heatmap without reordering genes
plotMarkerHeat(gs$inputMat, gs$inputClass, mm$markers)

#so now cluster genes and return the clustering: a better looking plot
pp = plotMarkerHeat(gs$inputMat, gs$inputClass, mm$markers, clusterGenes=TRUE, outs=TRUE)
pp$gene_class_info #cell clusters

#reorder cell clusters in the heatmap
plotMarkerHeat(gs$inputMat, gs$inputClass, mm$markers, clusterGenes=TRUE, newOrder = c(3,1,2))

#average every 4 cells to make a less intense heatmap
plotMarkerHeat(gs$inputMat, gs$inputClass, mm$markers, clusterGenes=TRUE, newOrder = c(3,1,2),
averageCells = 4)

#only cluster genes, do not make plots
pp = plotMarkerHeat(gs$inputMat, gs$inputClass, mm$markers, clusterGenes=TRUE, 
outs=TRUE, plotheat=FALSE)