ggplot: Try to plot boxplots with geom_rect on its background, but keep having error with object &qu-CodePudding

I was almost desperate with this error after working on this for 4 hrs, googled and looked from past posts already.

Here is my data structure:

str(tcga_exp)

'data.frame':   11775 obs. of  5 variables:
 $ cohort: chr  "BRCA-Basal.Tumor" "BRCA-LumA.Tumor" "BRCA-LumB.Tumor" "BRCA-LumA.Tumor" ...
 $ exp   : num  6.35 5.54 6.56 5.05 5.98 ...
 $ group : chr  "Tumor" "Tumor" "Tumor" "Tumor" ...
 $ comp  : chr  "no" "no" "no" "no" ...

But it won't work as this following code:

bp = ggplot(tcga_exp, aes(x=cohort, y=exp,color=group))  geom_rect(data = box_df,aes(xmin=normal,xmax=tumor,ymin=-Inf, ymax=Inf),color="grey",alpha= 0.2)  
  geom_boxplot()   geom_jitter(shape=16,alpha = 0.2,position=position_jitter(0.2),aes(color = group))  
   ylab(label = "log2(TPM 1)")   scale_color_manual(values = c("Tumor" = "#FD0606", "Normal" = "#1F25F9","Metastatic" = "#A432EB"))   
  theme_bw()   theme(axis.text.x = element_text(angle = 90,hjust=0.95,vjust=0.2),legend.position = "none",axis.title.x = element_blank())
# add signif
bp = bp   geom_signif(comparisons = compare,map_signif_level = T,tip_length = 0, color = "black",textsize = 3,vjust = 0.5)
bp

Error in FUN(X[[i]], ...) : object 'cohort' not found

But I think the "cohort" is in the column of my data.

But it works without the geom_rect part.

bp = ggplot(tcga_exp, aes(x = cohort, y = exp,color = group))   geom_boxplot()   geom_jitter(shape=16,alpha = 0.2,position=position_jitter(0.2),aes(color = group))   
   ylab(label = "log2(TPM 1)")   scale_color_manual(values = c("Tumor" = "#FD0606", "Normal" = "#1F25F9","Metastatic" = "#A432EB"))   
  theme_bw()   theme(axis.text.x = element_text(angle = 90,hjust=0.95,vjust=0.2),legend.position = "none",axis.title.x = element_blank())
# add signif
bp = bp   geom_signif(comparisons = compare,map_signif_level = T,tip_length = 0, color = "black",textsize = 3,vjust = 0.5)
bp

Supp:

str(box_df)
 $ normal: chr  "BLCA.Normal" "BRCA-Normal.Normal" "CHOL.Normal" "COAD.Normal" ...
 $ tumor : chr  "BLCA.Tumor" "BRCA.Tumor.Tumor" "CHOL.Tumor" "COAD.Tumor" ...

I am trying to make figures something like this enter image description here

Since the data is quite long, I uploaded the data to google drive for reproducibility. https://drive.google.com/file/d/1logQXXkmVpV0jy4i-MrMYRQ8UiFEOFmQ/view?usp=sharing

After keep working on this, I realized if you don't put aes() in the ggplot() but put behind: geom_boxplot and geom_jitter, you can plot the geom_rect, but that still doesn't solve my problem because then I can't do the geom_signif comparison.

bp = ggplot(tcga_exp)   geom_rect(data = box_df,aes(xmin = normal,xmax=tumor,ymin=-Inf, ymax=Inf),color="grey",alpha= 0.2)   geom_boxplot(aes(x = cohort, y = exp,color = group))   geom_jitter(shape=16,alpha = 0.2,position=position_jitter(0.2),aes(x = cohort, y = exp,color = group))   
   ylab(label = "log2(TPM 1)")   scale_color_manual(values = c("Tumor" = "#FD0606", "Normal" = "#1F25F9","Metastatic" = "#A432EB"))   
  theme_bw()   theme(axis.text.x = element_text(angle = 90,hjust=0.95,vjust=0.2),legend.position = "none",axis.title.x = element_blank())
# add signif
bp = bp   geom_signif(comparisons = compare,map_signif_level = T,tip_length = 0, color = "black",textsize = 3,vjust = 0.5)
bp
Error in f(...) : Can only handle data with groups that are plotted on the x-axis

Thanks!!!

CodePudding user response：

Map your primary aesthetics in ggplot() line. Then use inherit.aes=FALSE in the geom(s) that you want to define with different data.

dat <- readRDS("sample.RDS")
box_df <- data.frame(normal = c("BLCA.Normal", "BRCA-Normal.Normal", "CHOL.Normal", "COAD.Normal"), 
                  tumor = c("BLCA.Tumor", "BRCA.Tumor.Tumor", "CHOL.Tumor", "COAD.Tumor"))

ggplot(dat=dat, aes(x=cohort, y=exp, color=group))  
  geom_rect(data=box_df, aes(xmin=normal, xmax=tumor, ymin=-Inf, ymax=Inf),
            color = "grey", alpha=0.2, inherit.aes=FALSE)  
  geom_boxplot()  
  geom_jitter(shape=16, alpha=0.2, position=position_jitter(0.2))  
  geom_signif(comparisons=compare, map_signif_level=TRUE, tip_length=0, color="black",textsize=3, vjust=0.5, inherit.aes=FALSE)  
  scale_color_manual(values = c("Tumor" = "#FD0606", "Normal" = "#1F25F9","Metastatic" = "#A432EB"))  
  theme_bw()  
  theme(axis.text.x = element_text(angle=90, hjust=0.95, vjust=0.2),
        legend.position = "none", axis.title.x = element_blank())

I can't test this for completeness without all your data. If this doesn't work or you need more help, please let me know!