R/correspond_rank.R

In cola: A Framework for Consensus Partitioning

# == title
# Correspond two rankings
#
# == param
# -x1 A vector of scores calculated by one metric.
# -x2 A vector of scores calculated by another metric.
# -name1 Name of the first metric.
# -name2 Name of the second metric.
# -col1 Color for the first metric.
# -col2 Color for the second metric.
# -top_n Top n elements to show the correspondance.
# -transparency Transparency of the connecting lines.
# -pt_size Size of the points, must be a `grid::unit` object.
# -newpage Whether to plot in a new graphic page.
# -ratio Ratio of width of the left barplot, connection lines and right barplot. The three values will be scaled to a sum of 1.
# 
# == details
# In ``x1`` and ``x2``, the i^th element in both vectors corresponds to the same object (e.g. same row if they are calculated from a matrix) but with different 
# scores under different metrics.
# 
# ``x1`` and ``x2`` are sorted in the left panel and right panel respectively. The top n elements
# under corresponding metric are highlighted by vertical colored lines in both panels.
# The left and right panels also shown as barplots of the scores in the two metrics.
# Between the left and right panels, there are lines connecting the same element (e.g. i^th element in ``x1`` and ``x2``)
# in the two ordered vectors so that you can see how a same element has two different ranks in the two metrics.
#
# Under the plot is a simple Venn diagram showing the overlaps of the top n elements 
# by the two metrics.
#
# == value
# No value is returned.
#
# == seealso
# `correspond_between_rankings` draws for more than 2 sets of rankings.
#
# == author
# Zuguang Gu <z.gu@dkfz.de>
#
# == examples
# require(matrixStats)
# mat = matrix(runif(1000), ncol = 10)
# x1 = rowSds(mat)
# x2 = rowMads(mat)
# correspond_between_two_rankings(x1, x2, name1 = "SD", name2 = "MAD", top_n = 20)
correspond_between_two_rankings = function(x1, x2, name1, name2, 
	col1 = 2, col2 = 3, top_n = round(0.25*length(x1)), transparency = 0.9, 
	pt_size = unit(1, "mm"), newpage = TRUE, ratio = c(1, 1, 1)) {
	
	if(newpage) {
		grid.newpage()
	}

	if(length(x1) != length(x2)) {
		stop("Length of `x1` and `x2` should be the same.")
	}

	r1 = rank(x1, ties.method = "random")
	r2 = rank(x2, ties.method = "random")

	if(missing(name1)) name1 = deparse(substitute(x1))
	if(missing(name2)) name2 = deparse(substitute(x2))

	n = length(x1)
	text_height = grobHeight(textGrob("foo"))*2
	pushViewport(viewport(layout = grid.layout(nrow = 1, ncol = 3, widths = unit(ratio, "null")), 
		width = unit(1, "npc") - unit(2, "mm"),
		height = unit(1, "npc") - text_height - unit(1, "cm"), y = unit(1, "cm"), just = "bottom"))
	
	max_x1 = max(x1)
	pushViewport(viewport(layout.pos.row = 1, layout.pos.col = 1, 
		xscale = c(0, max_x1), yscale = c(0, n + 1)))
	grid.segments(max_x1 - x1, r1, max_x1, r1, default.units = "native", gp = gpar(col = "#EFEFEF"))
	l = r2 >= n - top_n
	grid.points(max_x1 - x1[l], r1[l], default.units = "native", pch = 16, size = pt_size, gp = gpar(col = add_transparency(col2, 0.5)))
	grid.text(name1, x = 1, y = unit(n + 1, "native") + unit(1, "mm"), default.units = "npc", just = c("right", "bottom"))
	upViewport()

	max_x2 = max(x2)
	pushViewport(viewport(layout.pos.row = 1, layout.pos.col = 3, 
		xscale = c(0, max_x2), yscale = c(0, n + 1)))
	grid.segments(0, r2, x2, r2, default.units = "native", gp = gpar(col = "#EFEFEF"))
	l = r1 >= n - top_n
	grid.points(x2[l], r2[l], default.units = "native", pch = 16, size = pt_size, gp = gpar(col = add_transparency(col1, 0.5)))
	grid.text(name2, x = 0, y = unit(n + 1, "native") + unit(1, "mm"), default.units = "native", just = c("left", "bottom"))
	upViewport()

	pushViewport(viewport(layout.pos.row = 1, layout.pos.col = 2, xscale = c(0, 1), yscale = c(0, n + 1)))
	l = r1 >= n - top_n | r2 >= n - top_n
	# if(sum(!l)) grid.segments(0, r1[!l], 1, r2[!l], default.units = "native", gp = gpar(col = "#EEEEEE80"))
	if(sum(l)) {
		grid.segments(0, r1[l], 1, r2[l], default.units = "native", gp = gpar(col = add_transparency("#000000", transparency)))
		# for(ind in which(l)) {
		# 	grid.bezier(c(0, 1, 0, 1), c(r1[ind], r1[ind], r2[ind], r2[ind]), default.units = "native", gp = gpar(col = add_transparency("#000000", transparency)))
		# }
	}
	grid.segments(c(0, 1), c(1, 1), c(0, 1), c(n - top_n, n - top_n), default.units = "native", gp = gpar(col = "#EEEEEE"))
	grid.segments(c(0, 1), c(n - top_n, n - top_n), c(0, 1), c(n, n), default.units = "native", gp = gpar(lwd = 4, col = c(col1, col2)))
	upViewport()

	upViewport()

	# add a venn diagram at the bottom
	n_intersect = length(intersect(order(x1, decreasing = TRUE)[1:top_n], order(x2, decreasing = TRUE)[1:top_n]))
	n_union = 2*top_n - n_intersect
	grid.roundrect(x = unit(0.5 - n_intersect/2/top_n*0.4, "npc"), y = unit(0.4, "cm"), width = unit(0.4, "npc"), 
		height = unit(0.4, "cm"), gp = gpar(fill = add_transparency(col2, 0.5), col = NA), just = "left")
	grid.roundrect(x = unit(0.5 + n_intersect/2/top_n*0.4, "npc"), y = unit(0.4, "cm"), width = unit(0.4, "npc"), 
		height = unit(0.4, "cm"), gp = gpar(fill = add_transparency(col1, 0.5), col = NA), just = "right")
	grid.text(qq("top @{top_n}/@{length(x1)}"), x = unit(0.5, "npc"), y = unit(0.7, "cm"), just = "bottom", gp = gpar(fontsize = 8))

}

# == title
# Correspond between a list of rankings
#
# == param
# -lt A list of scores under different metrics.
# -top_n Top n elements to show the correspondance.
# -col A vector of colors for ``lt``.
# -... Pass to `correspond_between_two_rankings`.
# 
# == details
# It makes plots for every pairwise comparison in ``lt``.
#
# == value
# No value is returned.
#
# == author
# Zuguang Gu <z.gu@dkfz.de>
# 
# == examples
# require(matrixStats)
# mat = matrix(runif(1000), ncol = 10)
# x1 = rowSds(mat)
# x2 = rowMads(mat)
# x3 = rowSds(mat)/rowMeans(mat)
# correspond_between_rankings(lt = list(SD = x1, MAD = x2, CV = x3), 
#     top_n = 20, col = c("red", "blue", "green"))
correspond_between_rankings = function(lt, top_n = length(lt[[1]]), 
	col = cola_opt$color_set_1[1:length(lt)], ...) {

	nm = names(lt)
	n = length(lt)
	n_plots = n*(n-1)/2

	if(length(col) == 1) {
		col = rep(col, n)
	}

	grid.newpage()
	pushViewport(viewport(layout = grid.layout(nrow = 1, ncol = n_plots)))
	k = 0
	for(i in seq_len(n-1)) {
		for(j in (i+1):n) {
			k = k + 1
			pushViewport(viewport(layout.pos.row = 1, layout.pos.col = k))
			pushViewport(viewport(width = 0.9))
			correspond_between_two_rankings(lt[[i]], lt[[j]], nm[i], nm[j], col[i], col[j], top_n, newpage = FALSE, ...)
			upViewport()
			upViewport()
		}
	}
	upViewport()
}