Color scales in the new version cannot be easily manipulated

[zuzkamat]

Hi,
I've been using clusterProfiler for a long time now and have been mostly satisfied with all the functionalities, thank you! However, in the recent update the color settings changed, which has been discussed already multiple times in the Issues here or elsewhere. I find the only possible solution via options() very inconvenient, because it prevents the color scales from being modified completely. I like to fine tune my scales through scale_color_gradient() and other settings, not only color-wise. This is impossible in the new version, as the options() overwrite everything else.

One recent example is the legend for p-value in a dotplot, which I would like to be horizontal and below the plot (see the code below). The default setting puts higher p-vals to the left and lower to the right. To improve readability of the plot, I would like to simply flip the colorbar with the values (lower left, higher right). It is impossible to do with standard settings like scale_color_gradient(low = "#990000", high = "#eaeaea", guide = guide_colorbar(reverse = TRUE)).

options(enrichplot.colours = c("#990000","#eaeaea"))
dotplot(merged.GO.plot, by = "Count", showCategory = 5, includeAll = T, label_format = 50) +
    theme(
          legend.position = "inside",
          legend.direction = "horizontal",
          legend.justification = c(1,-0.5),
          legend.box = "horizontal"
          ) 

Sure I'm not the only one fighting this or similar issues. Please, if there is any workaround, I would like to know, otherwise, could you consider this in the next update?

Thank you!

[guidohooiveld]

I know you posted your question a while ago, but it is (still!) possible to fine-tune plots! See code below.

Two remarks:
To adapt the color for the dotplot, the function scale_fill_gradient should be used (and not scale_color_gradient); thus fill vs scale). I also used reverse = FALSE.
To position the legend I used your code almost verbatim, except that to justify the legend I used the argument legend.justification.inside (note the addition of .inside to the function name); see: e.g. https://www.tidyverse.org/blog/2024/02/ggplot2-3-5-0-legends/#placement)

Using your use case with the included example data set:

> library(DOSE)
> library(clusterProfiler)
> library(enrichplot)
> library(ggplot2) ## explicitly loaded/required
> 
> ## using the example dataset
> data(gcSample)
> xx <- compareCluster(gcSample, fun="enrichGO", OrgDb="org.Hs.eg.db")
> xx <- pairwise_termsim(xx)
> 
> 
> ## create dotplot with default color scheme
> p1 <- dotplot(xx, by = "Count", showCategory = 5, includeAll = TRUE, label_format = 50)
> print(p1)
>

> ## change color scale
> p2 <- p1 + scale_fill_gradient(
+                  low = "#990000",
+                  high = "#eaeaea",
+                  guide = guide_colorbar(reverse = FALSE)
+                  )
Scale for fill is already present.
Adding another scale for fill, which will replace the existing scale.
>           
> 
> print(p2)
> 

> ## change position of legend
> ## values for legend.justification.inside represent location relative to x- resp y-axis
> p3 <- p2 +   theme(
+              legend.position = "inside",
+              legend.direction = "horizontal",
+              legend.justification.inside = c( 0.5, 0),
+              legend.box = "horizontal"
+              )
> 
> print(p3)
>

Note that legend is too big for the plot...
One solution is to make a bigger plot...

> dev.new(width = 350, height = 250, unit = "px")
> print(p3)
>

Alternatively, play with settings for argument:
legend.key.size, or legend.key.width together with legend.key.height.

Example:
legend.key.size = unit(30, 'pt')

or

legend.key.width = unit(20, "pt"),
legend.key.height = unit(10, "pt")

The info in this thread may provide some useful background info on the use of these functions: tidyverse/ggplot2#6433