Plotting violin plots: when I add a sample to be displayed, the violin plots no longer show up

Question

I am trying to display my results using violin plot and box plot at the same time.

I am using cell count to display the number of immune cells in different cancer samples/groups. When I plot the expression for 4 samples, everything works. When I add another sample (GTEx_M2), the violin plots for all other 4 samples disappear and I end up with only the box plots.

Any suggestion? Thanks in advance!

library(ggplot2)
library(ggpubr)
Cibersort7 = structure(list(
  Hot_M1 = c(0.0214400757119873, 0.170557805230298, 0.0804456569076382, 
             0.0893978598771954, 0.134477669028274, 0, 0.0525708788146097, 
             0.0511711964723951, 0.126904881120795, 0.0485101553521798, 
             0.170894800822398, 0.106555021195299, 0.0970104286070479, 
             0.115825265978309, 0.0427923320117795, 0.0733825856784013, 
             0.0111265771852828, 0.0657019859547462, 0.11656416302191,
             0.172002238486688, 0.0154591596631105, 0.0350445248592811, 
             0.0795539781894198, 0.0781276090630857, 0.0087982313041526, 
             0.0289274652853823, 0.0712661645666698, 0.0435482190581647, 
             0.0455556872660798, 0.0871522448556361), 
  Cold_M1 = c(0.0346024087291239, 0.0201947741817111, 0.0306194109725081, 
              0.0277445612030966, 0.00905915199266666, 0.00939058305405205, 
              0.0146535473252646, 0.0159980760737253, 0.147670469457772, 
              0.0426119074182886, 0.0219251208462312, 0.0128996237306264, 
              0.0094816829459359, 0.0219336027293415, 0.0438220246067735, 
              0.00950926112282649, 0.0838386603270565, 0.0486661009213444, 
              0.00651564872414969, 0.00110323590537234, 0.0807125087307139, 0, 
              0.037709808301658, 0, 0.0898041410439557, 0.0417739517920607, 0, 
              0.0202168551193018, 0.00176008746063679, 0.0161337603014608), 
  Hotnorm_M1 = c(0.00622155478760928, 0.00864956989565159, 0.0245812979257332, 
                 0.0339687958970202, 8e-04, 0, 0.0582086801600888, 0, 
                 0.03481918582501, 0.021338008027511, 0.0157360408231509, 
                 0.00489068636912568, 0.0281166183638247, 0.0162726467268935, 
                 0.0415769266772567, 0, 0.00344830695596762, 0.00196737745405557, 
                 0.0075141479562764, 0.0232464687737552, 0, 0, 0.0289423690350636, 
                 0.0218584208695064, 0.0255945495324721, 4e-04, 0.0221942067802419, 
                 0.00476738514342175, 0.00722699142988291, 0.00974645683928458), 
  Coldnorm_M1 = c(0.0280536098964266, 0.0261826834038114, 0.0150413750071331, 0, 
                  0.0199730743908202, 0.0115748800373456, 0.0275674859254823, 
                  0.0168847795974374, 0.0140281070945953, 0.00907861159279308, 
                  0, 0, 0, 0.0453414461512909, 0, 0.00730963773612433, 
                  0.0236424416792874, 0.0866914356225127, 0.0246339344582405, 
                  0.00881531992455549, 0.0140744199322424, 0, 0, 0, 
                  0.0319211626770028, 0.00155291355277603, 0.00295913497381517, 
                  0.00738775271575955, 0.0179786878323852, 0.00442919920031897), 
  GTEx_M1 = c(0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 
              0, 0, 0, 0.00551740159760184, 0, 0, 0, 0, 0)), 
  row.names = c(NA, -30L), 
  class = c("tbl_df", "tbl", "data.frame"))

This is a small part of my data that still shows the same issue I see.

y_axis  = list(na.omit(Cibersort7$Hot_M1), 
               na.omit(Cibersort7$Cold_M1), 
               na.omit(Cibersort7$Hotnorm_M1), 
               na.omit(Cibersort7$Coldnorm_M1), 
               na.omit(Cibersort7$GTEx_M1))
groupname = groupexpression = data = violinPlot  = pairwise_results = list(5)

for (i in 1:5){
  groupname[[i]] = as.factor(colnames(Cibersort7[, i]))
  groupexpression[[i]] = y_axis[[i]]
  data[[i]] = data.frame("Sample" = groupname[[i]], 
                         "Expression" = groupexpression[[i]])
}
dataframe = do.call(rbind, data)
dataframe$Sample = as.factor(dataframe$Sample)

my_comparisons = list(c("Hot_M1", "Cold_M1"),
                      c("Hot_M1", "Hotnorm_M1"), 
                      c("Hot_M1", "GTEx_M1"),
                      c("Cold_M1", "Coldnorm_M1"),
                      c("Cold_M1", "GTEx_M1"))

violinPlot = ggplot(dataframe, 
                    aes(x =Sample, y = Expression, fill = Sample)) + 
  geom_violin(trim = FALSE) + 
  geom_boxplot(width=0.1, fill="white") + 
  labs(title ="Distribution of M2 Macrophages", 
       x = "Tissue Samples", y = "Cibersort Count") + 
  theme_classic()

violinPlot

Here is how my violin plots look like:

Here is how they look like before adding the GTEx data:

And here's GTEx violin plots when displayed alone:

I understand that my GTEx data is zero but why do the violin plots disappear?

Answer 1

geom_violin has an argument named scale, which takes on the default value "area". From ?geom_violin:

if "area" (default), all violins have the same area (before trimming the tails). If "count", areas are scaled proportionally to the number of observations. If "width", all violins have the same maximum width.

Since GTEx's Expression values are concentrated at 0, its density peaks sharply at that value. We can see it more obviously in a normal density plot, with each sample's line overlaid atop one another:

ggplot(dataframe,
       aes(x = Expression, color = Sample)) +
  geom_density() +
  theme_classic()

With the default scale = "area" argument, including GTEx in the data means the violin plot for all other samples becomes a lot skinnier, & hence become almost completely covered by the boxplots. You'd still be able to see them if you comment out the boxplot layer.

You can set scale = "width" instead if you want comparable visibility between each violin. You may also want to highlight this to your target audience if you choose this option, as scale = "area" tends to be more common, & people may feel confused when some violins appear clearly larger than others.

ggplot(dataframe, 
       aes(x = Sample, y = Expression, fill = Sample)) + 
  geom_violin(trim = FALSE, scale = "width") +
  geom_boxplot(width=0.1, fill="white") +
  labs(title ="Distribution of M2 Macrophages", 
       x = "Tissue Samples", y = "Cibersort Count") + 
  theme_classic()

p.s. You can simplify your data processing steps, which are (from what I can tell) essentially a conversion from wide to long format. The usual way to do this is via melt (from reshape2 package) or gather (from tidyr package). Here's a possible implementation:

library(dplyr)
library(tidyr)

df2 <- Cibersort7 %>%
  gather(Sample, Expression) %>%
  mutate(Sample = factor(Sample, levels = colnames(Cibersort7)))

> all.equal(dataframe, as.data.frame(df2))
[1] TRUE

p.p.s. If there are multiple people commenting in your thread & you don't @ anyone in your reply, no one is going to get any notification about it, which is rather a waste if you've gone through all the trouble of improving your question. See here for an explanation of how the system works.