DataViz question for DEG and scRNA-seq

Question

[learner-MD]

Maybe a dumb question, but I recently came across a single-cell paper that showed "volcano" plots for all clusters in one image by filtering for all genes that were FDR <0.05, the x-axis being represented by different colors characterizing different clusters, the y-axis with positive log2FC going up and negative log2FC going down, and the significant genes plotted that way with select ones labeled.

I really liked that visual representation but did not save the article and can't find it again! If anyone either knows that article or has an R script approach to making such a visualization, I'd appreciate it greatly. Thanks in advance!

[Trivas]

Can you describe it a bit more? Is it like a volcano plot with the coordinates flipped and different "compartments" for the different clusters? Something like this (with dots instead of lines)?

[learner-MD]

Apologies for the delay and crude drawing, but was thinking something like the below (with all the plotted genes being FDR <0.05):

[Trivas]

One way to hack something together would be to plot your p-adj value on the x axis and log2FC on the y then adjust the p-adj values by intervals of .1 depending on your cluster to shift them to the right. You could overlay geom_rect() on the plot to be within 0-.1 on the x-axis and -.5-.5 on the y-axis for your labels. I hope someone else chimes in because this doesn't seem like an elegant solution.

df <- tibble(padj = runif(300, min=0, max=.1),
         l2fc = runif(300, min=-3, max = 3),
         cluster = c(rep(1, times = 100), rep(2, times = 100), rep(3, times = 100)))
df %<>% mutate(xmin = case_when(cluster == 1 ~ 0,
                                cluster == 2 ~ .1,
                                cluster == 3 ~ .2,),
               xmax = case_when(cluster == 1 ~ .1,
                                cluster == 2 ~ .2,
                                cluster == 3 ~ .3),
               ymin = -0.5,
               ymax = 0.5,
               padj_new = case_when(cluster == 1 ~ padj,
                                    cluster == 2 ~ padj+.1,
                                    cluster == 3 ~ padj+.2))

  ggplot(df, aes(x = padj_new, y=l2fc, color = cluster))+
      geom_point()+
      geom_rect(aes(xmin = xmin, xmax = xmax, ymin = ymin, ymax = ymax, fill = cluster))+
      geom_text(aes(x = xmin + (xmax - xmin)/2, y = ymin + (ymax - ymin)/2, label = paste0("Cluster: ", cluster)), 
                size = 10, color = "#FFFFFF")+
      theme_classic()

[Trivas]

Some personal changes: I would hide the entirety of the x-axis

  theme(axis.text.x = element_blank(),
        axis.line.x = element_blank(),
        axis.title.x = element_blank(),
        axis.ticks.x = element_blank())

I would also convert df$cluster to character before plotting.