Setting up custom color scale in R

Question

I am trying to setup a custom scale in R. My data ranges in values from -5.4 to +3.6 and I want to center the data around 0 (white). I would like a scale to the data such that I have the same number of gradations above and below 0 (I am shooting for 7 at the moment). The issue that I am having is that I can't get the scaling correct and I am not sure where my problem is.

My code at the moment (source data is in Pastebin link at the bottom):

   png('127-2_4_compare_other.png',width = 1200, height = 800, units = "px")
   colfunc <- colorRampPalette(c("blue", "white", "red"))
   f <- function(m) t(m)[,nrow(m):1]
   colorBarz=matrix(seq(-5.5,4,len=15),nrow=1)
   colorBarx=1
   source("127-2_4.CompareMatrix.txt")
   colorBary=seq(-5.4,3.6,len=15)
   cus_breaks=c(-5.400, -4.725, -4.050, -3.375, -2.700, -2.025, -1.350, -0.675,  0.45, 0.90, 1.35, 1.80, 2.25, 2.70, 3.15, 3.60)
   layout(matrix(c(1,2), 1, 2, byrow = TRUE), widths=c(9,1))
   image(f(Compare2and4),axes=FALSE,ylab="Amino acids",xlab="Position",main="Sample 2 vs. 4",col=colfunc(15),breaks=cus_breaks)
   axis(1, seq(from = 0, to = 1, by = 0.03703), labels=c(1:11,1:17))
   axis(2, seq(from = 0, to = 1, by = 0.0526),labels=rev(c("A","R","N","D","C","E","Q","G","H","I","L","K","M","F","P","S","T","W","Y","V")),las=2)
   image(colorBarx,colorBary,colorBarz,col=colfunc(15),axes=FALSE,xlab="",ylab="log(Sample4 / Sample2)",breaks=cus_breaks)
   axis(2,las=2)
   dev.off()

I am looking for seven evenly split bins above 0 to 3.6 and seven evenly split bins below 0 to -5.4 and I would like 0 to hit in the middle of the white bin. Also if anyone can look over the heatmap code itself to make sure that there are no obvious errors I would greatly appreciate it. Pastebin of the source data

Answer 1

A problem with your proposed color scale is that the perceptual distance between each point on the scale is not equal. The same color distance refers a larger range of data values on the positive end, compared to the negative end. I would suggest not reinventing the wheel, and making use of the great capabilities of ggplot.

By default, the diverging color scale will have white as the midpoint at value 0.

For example:

Compare2and4_t <- t(Compare2and4)[,nrow(Compare2and4):1]
am_ac <- c("A","R","N","D","C","E","Q","G","H","I","L","K","M","F","P","S","T","W","Y","V")
pdat <- data.frame(x = 1:nrow(Compare2and4_t), 
                   y = factor(rep(am_ac, each = nrow(Compare2and4_t)), levels = rev(am_ac)),
                   val = c(Compare2and4_t))

library(ggplot2)
library(scales)

ggplot(pdat, aes(x, y, fill = val)) + 
  geom_tile() +
  scale_fill_gradient2(low = 'darkblue', high = 'darkred', 
                       limits = c(-5.4, 3.6), oob = squish) +
  coord_equal(expand = FALSE) +
  scale_x_continuous(breaks = unique(pdat$x)) +
  theme_classic() +
  labs(x = 'Position', y = 'Amino acids')