mgheatmap: Creates a "geneset smart" ComplexHeatmap::Heatmap

In lianos/multiGSEA: A unified interface to a plethora of gene set enrichment analysis methods

Description

Encapsulates many common "moves" you'll make when trying to make a heatmap, especially if you are trying to show geneset activity across a panel of samples.

NOTE: this function will almost certainly reorder the rows of the input matrix. If you are concatentating Heatmap objects together horizontally (ie. you if you want to use a rowAnnotation along side the returned heatmap), you must reorder the rows of the annotation data.frame, ie. ranno.df <- ranno.df[rownames(out@matrix),]

Usage

mgheatmap(
  x,
  gdb = NULL,
  col = NULL,
  aggregate.by = c("none", "ewm", "ewz", "zscore"),
  split = TRUE,
  scores = NULL,
  gs.order = NULL,
  name = NULL,
  rm.collection.prefix = TRUE,
  rm.dups = FALSE,
  recenter = FALSE,
  rescale = FALSE,
  center = TRUE,
  scale = TRUE,
  rename.rows = NULL,
  zero_center_colramp = NULL,
  zlim = NULL,
  transpose = FALSE,
  ...
)

Arguments

x	the data matrix
gdb	GeneSetDb object that holds the genesets to plot. Defaults to NULL, which will plot all rows in x.
col	a colorRamp(2) function
aggregate.by	the method used to generate single-sample geneset scores. Default is none which plots heatmap at the gene level
split	introduce row-segmentation based on genesets or collections? Defaults is TRUE which will create split heatmaps based on collection if aggregate.by != 'none', or based on gene sets if aggregate.by == "none".
scores	If aggregate.by != "none" you can pass in a precomupted scoreSingleSamples() result, otherwise one will be computed internally. Note that if this is a data.frame of pre-computed scores, the gdb is largely irrelevant (but still required).
gs.order	This is experimental, and is here to help order the order of the genesets (or genesets collection) in a different way than the default. By default, gs.order = NULL and genesets are enumerated in alphabetical in the heatmap. You can pass in a character vector that will dictate the order of the genesets displayed in the heatmap. Currently this only matches against the "name" value of the geneset and probably only works when split = TRUE. We will support colleciton,name tuples soon. This can be a superset of the names found in gdb. As of ComplexHeatmap v2 (maybe earlier versions), this doesn't really work when cluster_rows = TRUE.
rm.dups	if aggregate.by == 'none', do we remove genes that appear in more than one geneset? Defaults to FALSE
recenter	do you want to mean center the rows of the heatmap matrix prior to calling ComplexHeatmap::Heatmap()?
rescale	do you want to standardize the row variance to one on the values of the heatmap matrix prior to calling ComplexHeatmap::Heatmap()?
rename.rows	defaults to NULL, which induces no action. Specifying a paramter here assumes you want to rename the rows of the heatmap. Please refer to the "Renaming Rows" section for details.
zero_center_colramp	Used to specify the type of color ramp to generate when col is NULL. By default (NULL) we try to guess if we should generate a 0-centered (blue, white, red) color ramp, or an absolute (viridis style) one. The guessing functionality isn't that great, so it doesn't hurt to explicitly set this to TRUE or FALSE.
zlim	Used to control the color saturation of the heatmap when the col parameter is not provided. If NULL, (default), extreme values (outside the c(0.025, 0.975) quantiles) are axed and the colorRamp is based on the remaining value range. If FALSE, the range of the colorRamp is defined by the min/max values. Otherwise a length(2) numeric can be supplied. If the values are between [0,1], then we assume this is a quantile range to be calculated. Otherwise the number are assumed to mark the top and bottom of the color scale range you want to use.
transpose	Flip display so that rows are columns. Default is FALSE.
...	parameters to send down to scoreSingleSamples() or ComplexHeatmap::Heatmap().

Details

More info here.

Value

A Heatmap object.

Renaming Heatmap Rows

This function leverages rename_rows() so that you can better customize the output of your heatmaps by tweaking its rownames.

If you are plotting a gene-level heatmap (ie. aggregate.by == "none"``) and the rownames()are gene identifiers, but you want the rownames of the heatmap to be gene symbols. You can perform this renaming using therename.rows' parameter.

• If rename.rows is NULL, then nothing is done.

• If rename.rows is a string, then we assume that x has an associated metadata data.frame over its rows and that rename.rows names one of its columns, ie. DGEList$genes[[rename.rows]] or fData(ExpressionSet)[[rename.rows]]. The values in that column will be swapped out for x's rownames

• If rename.rows is a two-column data.frame, the first column is assumed to be rownames(x) and the second is what you want to rename it to.

• When there are duplicates in the renamed rownames, the rename.duplicates ... parameter dictates the behavior. This will happen, for instance, if you are trying to rename the rows of an affy matrix to gene symbols, where we have multiple probe ids for one gene. When rename.duplicates is set to "original", one of the rows will get the new name, and the remaning duplicate rows will keep the rownames they came in with. When set to "make.unique", the new names will contain *.1, *.2, etc. suffixes, as you would get from using base::make.unique().

Maybe you are aggregating the expression scores into geneset scores, and you don't want the rownames of the heatmap to be collection;;name (or just name when rm.collection.prefx = TRUE), you can pass in a two column data.frame, where the first column is collection;name and the second is the name you want to rename that to. There is an example of this in the "Examples" section here.

Examples

library(ComplexHeatmap)
vm <- exampleExpressionSet()
gdb <- exampleGeneSetDb()
col.anno <- ComplexHeatmap::HeatmapAnnotation(
  df = vm$targets[, c("Cancer_Status", "PAM50subtype")],
  col = list(
    Cancer_Status = c(normal = "grey", tumor = "red"),
    PAM50subtype = c(Basal = "purple", Her2 = "green", LumA = "orange")))
mgh <- mgheatmap(vm, gdb, aggregate.by = "ewm", split=TRUE,
                 top_annotation = col.anno, show_column_names = FALSE,
                 column_title = "Gene Set Activity in BRCA subset")

# Maybe you want the rownames of the matrix to use spaces instead of "_"
rr <- geneSets(gdb)[, "name", drop = FALSE]
rr$newname <- gsub("_", " ", rr$name)
mg2 <- mgheatmap(vm, gdb, aggregate.by='ewm', split=TRUE,
                 top_annotation = col.anno, show_column_names = FALSE,
                 column_title = "Gene Set Activity in BRCA subset",
                 rename.rows = rr)

lianos/multiGSEA documentation built on Nov. 17, 2020, 1:26 p.m.