How to specify manual color scale in ggplot2 of R?

Question

I have ggplot with manual color scale by specify midpoint = 2.5 as below;

#create dataframe
df <-data.frame(x = c(rnorm(300, 3, 2.5), rnorm(150, 7, 2)), # create random data
                  y = c(rnorm(300, 6, 2.5), rnorm(150, 2, 2)),
                  z = c(rnorm(300, 6, 2.5), rnorm(150, 2, 2)))
#plot
gg <-ggplot(df, aes(x=x, y=y, color=z)) + 
       geom_point() + 
       scale_color_gradient2(midpoint=2.5, low="red", mid="white",high="blue", space ="Lab")

, which results in following figure;

Here, I would like to set more vivid (or deep) red color (ie #FF0000 or rgb(255,0,0)) at the minimum edge of the color scale (iemin z value is min(df$z)[1] -3.718939 ).

In this case, I do not want to move midpoint = 2.5 .

Do you have any solution?

Added note in response to the first answer

I want to keep blue vivid color at the z maximum as in the original figure. I have noticed the existence of trans = in scale_color_gradient2 thanks to the first answer. But, I have no idea how to solve my question.

Answer 1

Considering your sample, the mid point splits the data by 25% and 75% approximately.

Instead of having scale_color_gradient2 with three color calls, we can have scale_color_gradientn with four color calls and white as the second color (as the mid pint is just above 25%.

gg <-ggplot(df, aes(x=x, y=y, color=z)) + 
  geom_point() + 
  scale_color_gradientn(colors=c("red","white", "blue", "darkblue"), space ="Lab")

PS: You can also try colors=c("red","white", "lightblue", "blue")

Answer 2

Your gradient is already set with #FF0000 as the minimum. The scale is linear, so there's no getting around that.

Instead, you can consider transforming your color scale with trans= in the scale_color_gradient2 call.

The defaults are log , log2 and sqrt . None of these are terribly helpful due to having negative values for z .

You could instead set up a custom tranformation function with scales :

library(ggplot2)
library(scales)
asinh_trans <- scales::trans_new("asinh_trans",
                                transform=function(x) asinh(x),
                                inverse=function(x) sinh(x))

ggplot(df, aes(x=x, y=y, color=z)) + 
       geom_point() + 
       scale_color_gradient2(midpoint=asinh(2.5),
                             low="red",
                             mid="white",
                             high="blue",
                             trans=asinh_trans,
                             space ="Lab")

I'll agree that this doesn't look perfect, but you can easily set up any arbitrary transformation by setting your own trans_new function.