Create Legend for geom_text labels

Question

I would like to create a scatter plot were each point is graphed as a number instead of a circle. I need a legend that maps the number label in the scatter plot to the sample name.

My current attempt uses this code:

plot<-ggplot(df_mds,aes(x=x,y=y,group=interaction(Group,Sample)))+geom_text(aes(label=Label,color=Group,group=Sample),show.legend =T)
plot<-plot+guide()
plot<-plot+labs(x="Leading logFC dimension 1", y="Leading logFC dimension 2")

Using this data:

And creates this graph:

The graph above is almost exactly what I want except I need a legend mapping the values in the "Label" column to the values in the "Sample" column.

I have seen this post but I don't know how to combine the different shapes to create new numbers.

Any help would be greatly appreciated.

Answer 1

What you're looking for is the ability to create a legend based on the label= aesthetic. Unfortunately, the answer is that you cannot - at least not easily using ggplot . I would suggest using alternative approaches. Here I present two options using the following example dataset:

scatter.plot <- data.frame(
    sampleName=paste0('Sample_',1:18),
    x=sample(1:10, 18, replace=TRUE),
    y=sample(1:10, 18, replace=TRUE),
    id=1:18,
    grp=c(rep('Group A', 6), rep('Group B', 6), rep('Group C',6))
)

Use labels on your points and the ggrepel package to avoid overlapping labels.

Your sample names are short enough where it makes sense that you can add your labels on the plot itself. It can get a bit crowded, so I would recommend using the geom_text_repel and geom_label_repel functions in the ggrepel package to help with overplotting. Aesthetics and various parameters can be adjusted to your liking:

ggplot(scatter.plot, aes(x,y)) +
    geom_point(aes(color=grp), size=2.5) +
    geom_text_repel(
        aes(label=sampleName), color='gray45',
        min.segment.length = 0, size=3,
        force=10, direction='y') +
    theme_bw()

Use Color and Shape

Not sure if this would work for you, but if you adjust the shape for your group and shape for the sample, you can get a legend for each sample out of the box using ggplot fairly easily. Of course, it needs color to be able to discriminate properly:

ggplot(scatter.plot, aes(x,y)) +
    geom_point(aes(shape=grp, color=sampleName), size=2)