Boxplots for the same columns in different data frames

Question

I have 5 data frames that contain the prediction scores of ML models. Those data frames have the same exact columns, just different samples (each data frame was made with a different seed). Lets call the data frames seed1, seed2 ...

There are 3 types of columns that I'm interested in. Columns that start with cr, with pd and with st_po. In each one of the 5 data frames there are 4 columns that start with cr, 4 that start with pd and 4 that start with st_po.

I want to make boxplots with rows. Each row represents a data frame and its 4 columns. Overall I'll have 3 plots. A plot for the cr columns, a plot for the pd columns and a plot for the st_po.

For example the cr plot would look something like this, with 5 rows, in each row there are 4 boxes (cause of the 4 columns, each color represents a column):

Data - this is a subset, for one of the five data frames. The other 4 look exactly the same just different sample names.

structure(list(st_po_meta = c(0.382992297410965, 0.460950464010239, 
0.447804838418961, 0.447804838418961, 0.460950464010239, 0.447804838418961, 
0.369836807250977, 0.447804838418961, 0.369836807250977, 0.447804838418961
), st_po_meta_added = c(0.460011065006256, 0.52004611492157, 0.253930300474167, 
0.222006008028984, 0.302200853824615, 0.485153168439865, 0.20485857129097, 
0.350892871618271, 0.331338971853256, 0.295754462480545), st_po_meta_genes = c(0.277256995439529, 
0.425392180681229, 0.182383552193642, 0.253527283668518, 0.329186052083969, 
0.305586904287338, 0.188975885510445, 0.238625407218933, 0.497761845588684, 
0.342641144990921), st_po_all = c(0.565486133098602, 
0.564990341663361, 0.164183273911476, 0.15946152806282, 0.234778091311455, 
0.396436214447021, 0.172556579113007, 0.257463246583939, 0.43759897351265, 
0.200696632266045), cr_meta = c(0.0779446139931679, 0.274154871702194, 
0.0718425810337067, 0.0718425810337067, 0.274154871702194, 0.0718425810337067, 
0.160841777920723, 0.0718425810337067, 0.160841777920723, 0.0718425810337067
), cr_meta_added = c(0.0662130266427994, 0.258626192808151, 0.0601647943258286, 
0.060332003980875, 0.141631454229355, 0.113691322505474, 0.0786028951406479, 
0.0668068528175354, 0.171470999717712, 0.0754544585943222), cr_meta_genes = c(0.144175127148628, 
0.145591989159584, 0.0984272509813309, 0.0906868129968643, 0.28544145822525, 
0.138114541769028, 0.091837003827095, 0.0904595032334328, 0.211963757872581, 
0.163982316851616), cr_all = c(0.118267595767975, 
0.199180424213409, 0.0867970511317253, 0.0653180256485939, 0.203389659523964, 
0.149213299155235, 0.126333728432655, 0.0975232273340225, 0.256154090166092, 
0.146669581532478), pd_meta = c(0.287342727184296, 0.417357206344604, 
0.255898356437683, 0.255898356437683, 0.417357206344604, 0.255898356437683, 
0.471998482942581, 0.255898356437683, 0.471998482942581, 0.255898356437683
), pd_meta_added = c(0.299634903669357, 0.418549239635468, 0.424916654825211, 
0.432371437549591, 0.584436595439911, 0.296403467655182, 0.563782930374146, 
0.3688924908638, 0.578023612499237, 0.278784334659576), pd_meta_genes = c(0.314525783061981, 
0.343217849731445, 0.315391361713409, 0.353350460529327, 0.562197327613831, 
0.292534917593002, 0.616392850875854, 0.284660279750824, 0.532478809356689, 
0.341239869594574), pd_all = c(0.216887220740318, 
0.283130913972855, 0.248720198869705, 0.421182304620743, 0.564142644405365, 
0.204288363456726, 0.778401911258698, 0.272624760866165, 0.511143803596497, 
0.186424404382706)), row.names = c("04d83340b8bd", "122T", "1c2a5ac94492", 
"1d209304d988", "212T", "24ab7fecc92e", "356T", "379fe8924c51", 
"39T", "3ec4d3fc8bd1"), class = "data.frame")

Answer 1

Reformat the data before plotting, here we are row binding all the seeds, as an example I am using the same dataframe 3 times - 1:3 seeds. Then subset by column types and convert wide-to-long, then rowbind all again. Finally, we have a long format data suitable for ggplot using facets:

# create example 3 dataframes
seed2 <- seed3 <- seed1

library(ggplot2)
library(data.table)

d <- rbindlist(mget(ls(pattern = "^seed")), idcol = "data")

dplot <- rbindlist(
  sapply(c("cr", "pd", "st_po"), function(i){
    cols <- c("data", colnames(d)[ startsWith(colnames(d), i) ])
    x <- melt(d[, ..cols ], id.vars = "data", variable.name = "grp", value.name = "value")
    x[,  grp := gsub(paste0(i, "_"), "", grp)]
    x}, simplify = FALSE),
  idcol = "type")
  
ggplot(dplot, aes(grp, value, fill = grp)) +
  geom_boxplot() +
  facet_grid(type~data)

Or flip the facet:

ggplot(dplot, aes(grp, value, fill = grp)) +
  geom_boxplot() +
  facet_grid(data~type)