How do I assign 9 columns specific colours of independent gradient on my heatmap?

Question

I have a file which contains the following snippet except with 3000 entries of different animals and blood type

FileA

Animal Bloodtype Count
Horse  Opos      10
Horse  Apos      5
Horse  Bpos      4
Horse  ABpos     5
Horse  Oneg      6
Horse  Aneg      7
Horse  Bneg      9
Horse  ABneg     10
Horse  Unknown   10
Cat    Opos      12
Cat    Apos      15
Cat    Bpos      14
Cat    ABpos     15
Cat    Oneg      16
Cat    Aneg      17
Cat    Bneg      19
Cat    ABneg     14
Cat    Unknown   14
Dog    Opos      9
Dog    Apos      23
Dog    Bpos      12
Dog    ABpos     42
Dog    Oneg      45
Dog    Aneg      23
Dog    Bneg      45
Dog    ABneg     32
Dog    Unknown   32
Mouse  Opos      3
Mouse  Apos      4
Mouse  Bpos      5
Mouse  ABpos     3
Mouse  Oneg      6
Mouse  Aneg      8
Mouse  Bneg      8
Mouse  ABneg     20
Mouse  Unknown   20
Pig    Opos      19
Pig    Apos      13
Pig    Bpos      22
Pig    ABpos     32
Pig    Oneg      25
Pig    Aneg      13
Pig    Bneg      35
Pig    ABneg     22
Pig    Unknown   22

I am trying to produce a heatmap where my y-axis are the animals, the bloodtype on x-axis and the counts as values.

I am trying to color each column independently by bloodtype with its own specific colour and ascending gradient per column to easily tell what animal have high numbers of O-positive, or A-positive etc. and what animals are running low via decreasing gradient..etc (because the bloodtypes are color-coded for easy visualisation)

Basically, I have tried to do something like what was done in this stackoverflow question: ggplot2 heatmaps: using different gradients for categories

or this one but different colours per row: Heat map per column with ggplot2

csv_file<-read.csv("~/Documents/FileA.csv")

csv_file.s <- ddply(csv_file, .(Bloodtype), transform, rescale = scale(Count))

csv_file.s$Category <- csv_file.s$Bloodtype

levels(csv_file.s$Category) <- 
  list("Opos" = c("Opos"),
       "Apos" = c("Apos"),
       "Bpos" = c("Bpos"),
       "ABpos" = c("ABpos"),
       "Oneg" = c("Oneg"),
       "Aneg" = c("Aneg"),
       "Bneg" = c("Bneg"),
       "Oneg" = c("Oneg"),
       "Unknown" = c("Unknown"))

csv_file.s$rescaleoffset <- csv_file.s$rescale + 100*(as.numeric(as.factor(csv_file.s$Category))-1)
scalerange <- range(csv_file.s$rescale)
gradientends <- scalerange + rep(c(0,100,200), each=8)
colorends <- c("white", "Aquamarine4", "white", "yellow4", "white", "turquoise4","white","orange4", "white", "slategray4","white","seagreen4","white","purple4","white","red4","white","blue4")


ggplot(csv_file.s, aes(Bloodtype, Animal)) + 
  geom_tile(aes(fill = rescaleoffset), colour = "transparent") + 
  scale_fill_gradientn(colours = colorends, 
                       values = rescale(gradientends)) + 
  scale_x_discrete("", expand = c(0, 0))+ 
  scale_y_discrete("", expand = c(0, 0)) +
  theme(panel.background = element_rect(fill = 'white'))
  theme_grey(base_size = 12) + 
  theme(legend.position = "none",
        axis.ticks = element_blank(), 
        axis.text.x = element_text(angle = 330, hjust = 0))

But the gradient turns out wrong and the colours are all over the place. I've been trying to find how to assign colours to specific column headers in heatmap, i.e Unknown="blue4", ABneg="red4", but to no avail. Basically, I don't know what I'm doing. :(

Any help would be greatly appreciated.

score 2 · Accepted Answer · answered May 21 '19 at 21:04

2

So the easiest solution I could think of is to simply map the bloodtype to a fill colour and map the count to an alpha scale, such that high counts have high colour intensity and low counts are near-white. I'm not sure what you meant with ascending and descending gradients, so I mostly ignored that.

Assume df is your snippet of data in data.frame format:

ggplot(df, aes(x = Bloodtype, y = Animal, fill = Bloodtype, alpha = Count)) +
  # Dummy tile geom for white background
  geom_tile(fill = "white", alpha = 1) +
  geom_tile() +
  scale_alpha_continuous(breaks = seq(0, max(df$Count), length.out = 10), 
                         limits = c(0, NA))

You may have to fiddle around a bit with the breaks and limits in the alpha scale to match your data. Of course, you can choose any colours you want for the fill by adding a scale_fill_*() to the plot.

answered May 21 '19 at 21:04

teunbrand

33,645
4
37
63

Thanks teunbrand! I did scale_fill_manual for assigning colours and it worked! Would you happen to know how I can hide just the "Count" legend? – cms72 May 22 '19 at 02:23
Yes you could add `guide = "none"` to the `scale_alpha_continuous()` arguments. Glad it worked out! – teunbrand May 22 '19 at 06:24
Thank you! Just one more question, would you happen to know how to make all the colours a shade darker. What I mean to say is I have a range of numbers between 1-300, because of that, counts with 1 are barely visible - but I would still like to distinguish them from a count of 300. – cms72 May 23 '19 at 03:44
You could try playing around with `limits = c(..., ...)` inside the `scale_alpha_continuous()`, to get the limits to start at some number (for example `c(-5, NA)` to have the counts = 1 be somewhat more coloured). An other option is to transform your counts with for example `log10()`, `sqrt()` or `asinh()`, to highlight that the differense between 1 and 5 is more important than the difference between 41 and 45. – teunbrand May 23 '19 at 06:42
Ohh is that what limits does. Great ideas. I'll give them a shot. Thank you so much @teunbrand! – cms72 May 23 '19 at 08:16

How do I assign 9 columns specific colours of independent gradient on my heatmap?

1 Answers1