25

If any, I think there must be a very easy solution for this. I have two large dataframes that basically look like these:

> data1[1,]
      chromosome start    end      test ref position log2      p.value 
13600 Y          10199251 10200750 533  616 10200000 0.2181711 0.00175895   
...

> data2[1,]
      chromosome start    end      test ref position log2       p.value 
4080  Y          10197501 10202500 403  367 10200000 0.04113596 0.3149926   
...

I'm using this code to plot the two dataframes into the same graph:

p <- ggplot() + geom_point(data=subset(data1, p.value >= glim[1]),
map=aes(x=position, y=log2, colour=p.value))
+ geom_point(data=subset(data2, p.value >= glim[1]), map=aes(x=position,
y=log2, colour=p.value))

When I was plotting single dataframes, I was using a red-white color gradient for the values in "p.value" column. Using this line:

p <- p + scale_colour_gradient(limits=glim, trans='log10', low="red", 
high="white")

The central issue is: Now with two dataframes, how can I set one color gradient for data1 and another for data2? I read in a previous post that it is not possible to use two different colour scales(ej. "low=" for the first, and "high=" for the second), but in this case is exactly the same kind of colour scale (If I'm not mixing up terminology). The syntax obviously is not correct but I'd like to do something like this:

p <- p + scale_colour_gradient(limits=glim, trans='log10', low="red", 
high="white") 

p <- p + scale_colour_gradient(limits=glim, trans='log10', low="blue", 
high="white")
Julius Vainora
  • 47,421
  • 9
  • 90
  • 102
JRodrigoF
  • 855
  • 1
  • 11
  • 20
  • [Relevant post](http://stackoverflow.com/a/5738083/680068) uses `scale_colour_identity()` to define your own colours for each point. – zx8754 Aug 08 '14 at 08:55

2 Answers2

28

I know this is a long time after, but have a look at https://github.com/eliocamp/ggnewscale.

You can use the function new_colour_scale() to start a new colour scale that applies to everything after it.

masher
  • 3,814
  • 4
  • 31
  • 35
26

First, note that the reason ggplot doesn't encourage this is because the plots tend to be difficult to interpret.

You can get your two color gradient scales, by resorting to a bit of a cheat. In geom_point certain shapes (21 to 25) can have both a fill and a color. You can exploit that to create one layer with a "fill" scale and another with a "color" scale.

# dummy up data
dat1<-data.frame(log2=rnorm(50), p.value= runif(50))
dat2<-data.frame(log2=rnorm(50), p.value= runif(50))

# geom_point with two scales
p <- ggplot() +
       geom_point(data=dat1, aes(x=p.value, y=log2, color=p.value), shape=21, size=3) +
       scale_color_gradient(low="red", high="gray50") +
       geom_point(data= dat2, aes(x=p.value, y=log2, shape=shp, fill=p.value), shape=21, size=2) +
       scale_fill_gradient(low="gray90", high="blue")
p

enter image description here

tjebo
  • 21,977
  • 7
  • 58
  • 94
Ram Narasimhan
  • 22,341
  • 5
  • 49
  • 55
  • Aha, I am having the same problem except that I am not applying scale_colour_gradient to two sets of points, but to a single set of points, and I would like another scale_colour_gradient for a superimposed geom_path..... both have differing distributions but I still would like to see them in the same chart. Any hints? I should add that my gradients are heavily customized using scale_colour_gradientn – Thomas Browne May 18 '14 at 22:06
  • 1
    @ThomasBrowne If you can ask it as a new question, with a small reproducible example, there are many ggplot expert users here who will jump in and answer. – Ram Narasimhan May 19 '14 at 02:58
  • Sorry to bother, Ram! Would you mind helping me with my own answer to this question: http://stackoverflow.com/a/25588144/2872891? I also have another issue when integrating this approach in my real codebase. Could I ask you for help, when I will document it as a question? Thank you! – Aleksandr Blekh Aug 31 '14 at 17:57
  • A more general approach is to rescale your data to a common scale before calling ggplot2. See https://stackoverflow.com/a/5738083/3980197 – CoderGuy123 Jun 29 '17 at 21:13
  • 1
    what is `shp` here? in `shape = shp` – AnilGoyal Dec 23 '20 at 07:01