plotExprHeatmap: Plot expression heatmap

In HelenaLC/CATALYST: Cytometry dATa anALYSis Tools

Plot expression heatmap

Description

Heatmap of marker expressions aggregated by sample, cluster, or both; with options to include annotation of cell metadata factors, clustering(s), as well as relative and absolute cell counts.

Usage

plotExprHeatmap(
  x,
  features = NULL,
  by = c("sample_id", "cluster_id", "both"),
  k = "meta20",
  m = NULL,
  assay = "exprs",
  fun = c("median", "mean", "sum"),
  scale = c("first", "last", "never"),
  q = 0.01,
  row_anno = TRUE,
  col_anno = TRUE,
  row_clust = TRUE,
  col_clust = TRUE,
  row_dend = TRUE,
  col_dend = TRUE,
  bars = FALSE,
  perc = FALSE,
  bin_anno = FALSE,
  hm_pal = rev(brewer.pal(11, "RdYlBu")),
  k_pal = .cluster_cols,
  m_pal = k_pal,
  distance = c("euclidean", "maximum", "manhattan", "canberra", "binary", "minkowski"),
  linkage = c("average", "ward.D", "single", "complete", "mcquitty", "median",
    "centroid", "ward.D2")
)

Arguments

x	a SingleCellExperiment.
features	character string specifying which features to include; valid values are "type"/"state" for type/state_markers(x) if rowData(x)$marker_class have been specified; a subset of rownames(x); NULL to use all features. When by = "both", only 1 feature is allowed.
by	character string specifying whether to aggregate by sample, cluster, both.
k	character string specifying which clustering to use when by != "sample_id"; assay data will be aggregated across these cluster IDs.
m	character string specifying a metaclustering to include as an annotation when by != "sample_id" and row_anno = TRUE.
assay	character string specifying which assay data to use; valid values are assayNames(x).
fun	character string specifying the function to use as summary statistic.
scale	character string specifying the scaling strategy: "first": scale & trim then aggregate "last": aggregate then scale & trim "never": aggregate only If scale != "never", data will be scaled using lower (q%) and upper (1-q%) quantiles as boundaries.
q	single numeric in [0,0.5) determining the quantiles to trim when scale != "never".
row_anno, col_anno	logical specifying whether to include row/column annotations (see details); when one axis corresponds to samples (by != "cluster_id"), this can be a character vector specifying a subset of names(colData(x)) to be included as annotations.
row_clust, col_clust	logical specifying whether rows/columns should be hierarchically clustered and re-ordered accordingly.
row_dend, col_dend	logical specifying whether to include the row/column dendrograms.
bars	logical specifying whether to include a barplot of cell counts per cluster as a right-hand side row annotation.
perc	logical specifying whether to display percentage labels next to bars when bars = TRUE.
bin_anno	logical specifying whether to display values inside bins.
hm_pal	character vector of colors to interpolate for the heatmap.
k_pal, m_pal	character vector of colors to interpolate for cluster annotations when by != "sample_id".
distance	character string specifying the distance metric to use for both row and column hierarchical clustering; passed to Heatmap
linkage	character string specifying the agglomeration method to use for both row and column hierarchical clustering; passed to Heatmap

Details

By default (row/col_anno = TRUE), for axes corresponding to samples (y-axis for by = "sample_id" and x-axis for by = "both"), annotations will be drawn for all non-numeric cell metadata variables. Alternatively, a specific subset of annotations can be included for only a subset of variables by specifying row/col_anno to be a character vector in names(colData(x)) (see examples).

For axes corresponding to clusters (y-axis for by = "cluster_id" and "both"), annotations will be drawn for the specified clustering(s) (arguments k and m).

Value

a Heatmap-class object.

Author(s)

Helena L Crowell [e-mail]

References

Nowicka M, Krieg C, Crowell HL, Weber LM et al. CyTOF workflow: Differential discovery in high-throughput high-dimensional cytometry datasets. F1000Research 2017, 6:748 (doi: 10.12688/f1000research.11622.1)

See Also

plotFreqHeatmap, plotMultiHeatmap

Examples

data(PBMC_fs, PBMC_panel, PBMC_md)
sce <- prepData(PBMC_fs, PBMC_panel, PBMC_md)
sce <- cluster(sce) 

# median scaled & trimmed expression by cluster
plotExprHeatmap(sce, 
  by = "cluster_id", k = "meta8",
  scale = "first", q = 0.05, bars = FALSE)

# scale each marker between 0 and 1 
# after aggregation (without trimming)
plotExprHeatmap(sce, 
  scale = "last", q = 0,
  bars = TRUE, perc = TRUE,
  hm_pal = hcl.colors(10, "YlGnBu", rev = TRUE))

# raw (un-scaled) median expression by cluster-sample
plotExprHeatmap(sce,
  features = "pp38", by = "both", k = "meta10", 
  scale = "never", row_anno = FALSE, bars = FALSE)
  
# include only subset of samples
sub <- filterSCE(sce, 
  patient_id != "Patient",
  sample_id != "Ref3")
 
# includes specific annotations &
# split into CDx & all other markers
is_cd <- grepl("CD", rownames(sce))
plotExprHeatmap(sub, 
  rownames(sce)[is_cd], 
  row_anno = "condition", 
  bars = FALSE)
plotExprHeatmap(sub, 
  rownames(sce)[!is_cd], 
  row_anno = "patient_id",
  bars = FALSE)

HelenaLC/CATALYST documentation built on Nov. 30, 2024, 4:04 a.m.