Learn R Programming

seriation (version 1.5.6)

hmap: Plot Heat Map Reordered Using Seriation

Description

Provides heatmaps reordered using several different seriation methods. This includes dendrogram based reordering with optimal leaf order and matrix seriation-based heat maps.

Usage

hmap(
  x,
  distfun = stats::dist,
  method = "OLO_complete",
  control = NULL,
  scale = c("none", "row", "column"),
  plot_margins = "auto",
  col = NULL,
  col_dist = grays(power = 2),
  row_labels = NULL,
  col_labels = NULL,
  ...
)

gghmap( x, distfun = stats::dist, method = "OLO_complete", control = NULL, scale = c("none", "row", "column"), prop = FALSE, ... )

Value

An invisible list with elements:

rowInd, colInd

index permutation vectors.

reorder_method

name of the method used to reorder the matrix.

The list may contain additional elements (dendrograms, colors, etc).

Arguments

x

a matrix or a dissimilarity matrix of class dist. If a dissimilarity matrix is used, then the distfun is ignored.

distfun

function used to compute the distance (dissimilarity) between both rows and columns. For gghmap(), this parameter is passed on in control.

method

a character strings indicating the used seriation algorithm (see seriate.dist()). If the method results in a dendrogram then stats::heatmap() is used to show the dendrograms, otherwise reordered distance matrices are shown instead.

control

a list of control options passed on to the seriation algorithm specified in method.

scale

character indicating if the values should be centered and scaled in either the row direction or the column direction, or none. Default is none.

plot_margins

character indicating what to show in the margins. Options are: "auto", "dendrogram", "distances", or "none".

col

a list of colors used.

col_dist

colors used for displaying distances.

row_labels, col_labels

a logical indicating if row and column labels in x should be displayed. If NULL then labels are displayed if the x contains the appropriate dimname and the number of labels is 25 or less. A character vector of the appropriate length with labels can also be supplied.

...

further arguments passed on to stats::heatmap().

prop

logical; change the aspect ratio so cells in the image have a equal width and height.

Author

Michael Hahsler

Details

For dendrogram based heat maps, the arguments are passed on to stats::heatmap() in stats. The following arguments for heatmap() cannot be used: margins, Rowv, Colv, hclustfun, reorderfun.

For seriation-based heat maps further arguments include:

  • gp an object of class gpar containing graphical parameters (see gpar() in package grid).

  • newpage a logical indicating whether to start plot on a new page

  • prop a logical indicating whether the height and width of x should be plotted proportional to its dimensions.

  • showdist Display seriated dissimilarity matrices? Values are "none", "both", "rows" or "columns".

  • key logical; show a colorkey?

  • key.lab Label plotted next to the color key.

  • margins bottom and right-hand-side margins are calculated automatically or can be specifies as a vector of two numbers (in lines).

  • zlim range of values displayed.

  • col, col_dist color palettes used.

See Also

Other plots: VAT(), bertinplot(), dissplot(), palette(), pimage()

Examples

Run this code
data("Wood")

# regular heatmap from package stats
heatmap(Wood, main = "Wood (standard heatmap)")

# Default heatmap does Euclidean distance, hierarchical clustering with
# complete-link and optimal leaf ordering. Note that the rows are
# ordered top-down in the seriation order (stats::heatmap orders in reverse)
hmap(Wood, main = "Wood (opt. leaf ordering)")
hmap(Wood, plot_margins = "distances", main = "Wood (opt. leaf ordering)")
hmap(Wood, plot_margins = "none", main = "Wood (opt. leaf ordering)")

# Heatmap with correlation-based distance, green-red color (greenred is
# predefined) and optimal leaf ordering and no row label
dist_cor <- function(x) as.dist(sqrt(1 - cor(t(x))))
hmap(Wood, distfun = dist_cor, col = greenred(100),
  main = "Wood (reorded by corr. between obs.)")

# Heatmap with order based on the angle in two-dimensional MDS space.
hmap(Wood, method = "MDS_angle", col = greenred(100), row_labels = FALSE,
  main = "Wood (reorderd using ange in MDS space)")

# Heatmap for distances
d <- dist(Wood)
hmap(d, main = "Wood (Euclidean distances)")

# order-based with dissimilarity matrices
hmap(Wood, method = "MDS_angle",
  col = greenred(100), col_dist = greens(100, power = 2),
  keylab = "norm. Expression", main = "Wood (reorderd with distances)")

# without the distance matrices
hmap(Wood, method = "MDS_angle", plot_margins = "none",
  col = greenred(100), main = "Wood (reorderd without distances)")

# Manually create a simple heatmap with pimage.
o <- seriate(Wood, method = "heatmap",
   control = list(dist_fun = dist, seriation_method = "OLO_ward"))
o

pimage(Wood, o)

# Note: method heatmap calculates reorderd hclust objects which can be used
#       for many heatmap implementations like the standard implementation in
#       package stats.
heatmap(Wood, Rowv = as.dendrogram(o[[1]]), Colv = as.dendrogram(o[[2]]))

# ggplot 2 version does not support dendrograms in the margin (for now)
if (require("ggplot2")) {
  library("ggplot2")

  gghmap(Wood) + labs(title = "Wood", subtitle = "Optimal leaf ordering")

  # More parameters (see ? ggpimage): reverse column order and flip axes, make a proportional plot
  gghmap(Wood, reverse_columns = TRUE) +
    labs(title = "Wood", subtitle = "Optimal leaf ordering")

  gghmap(Wood, flip_axes = TRUE) +
    labs(title = "Wood", subtitle = "Optimal leaf ordering")

  gghmap(Wood, flip_axes = TRUE, prop = TRUE) +
    labs(title = "Wood", subtitle = "Optimal leaf ordering")

  dist_cor <- function(x) as.dist(sqrt(1 - cor(t(x))))
  gghmap(Wood, distfun = dist_cor) +
    labs(title = "Wood", subtitle = "Reorded by correlation between observations") +
    scale_fill_gradient2(low = "darkgreen", high = "red")

  gghmap(d, prop = TRUE) +
    labs(title = "Wood", subtitle = "Euclidean distances, reordered")

  # Note: the ggplot2-based version currently cannot show distance matrices
  #      in the same plot.

  # Manually seriate and plot as pimage.
  o <- seriate(Wood, method = "heatmap", control = list(dist_fun = dist,
    seriation_method = "OLO_ward"))
  o

  ggpimage(Wood, o)
}

Run the code above in your browser using DataLab