R-组合冲积图

Question

I have been wondering how one could generate an alluvial plot using a data frame in R (eg utilizing ggalluvial package)

I have a data frame like following

   GENE UNTREATED TREATED
1    G1      FOUR    FOUR
2    G2      FOUR    FOUR
3    G3      FOUR    FOUR
4    G4      FOUR   THREE
5    G5     THREE    NONE
6    G6     THREE    NONE
7    G7       TWO   THREE
8    G8       TWO     ONE
9    G9       ONE     ONE
10  G10       TWO     ONE

I would like to generate an alluvial plot with this data. For example, 2 bars in the plot will be UNTREATED and TREATED and connections will be based on the proportions of how genes changing their values before and after treatment. It should definitely possible to do this and I didn't seem to get the grasp of alluvial plot concept. Any help would be highly appreciated.

EDIT: Here is a sample sketch I made how the plot would look like

Thank you.

Answer 1

This seems to be what you are after

dd<-read.table(text="   GENE UNTREATED TREATED
1    G1      FOUR    FOUR
2    G2      FOUR    FOUR
3    G3      FOUR    FOUR
4    G4      FOUR   THREE
5    G5     THREE    NONE
6    G6     THREE    NONE
7    G7       TWO   THREE
8    G8       TWO     ONE
9    G9       ONE     ONE
10  G10       TWO     ONE", header=TRUE)

ggplot(dd, aes(axis1=UNTREATED, axis2=TREATED)) +
  geom_alluvium() + 
  geom_stratum(width = 1/12, fill = "black", color = "grey") +
  geom_label(stat = "stratum", label.strata = TRUE)

Answer 2

Alternatively you can use the ggalluvial wrappers that come with easyalluvial

 dd<-read.table(text="   GENE UNTREATED TREATED
               1    G1      FOUR    FOUR
               2    G2      FOUR    FOUR
               3    G3      FOUR    FOUR
               4    G4      FOUR   THREE
               5    G5     THREE    NONE
               6    G6     THREE    NONE
               7    G7       TWO   THREE
               8    G8       TWO     ONE
               9    G9       ONE     ONE
               10  G10       TWO     ONE", header=TRUE)


easyalluvial::alluvial_wide( dd, id = GENE, fill_by = 'all_flows',
                             order_levels = c('NONE','FOUR', 'THREE', 'TWO', 'ONE'))

Answer 3

MrFlick answer is very nice. Another option would be using alluvial plots, which has been discussed in " https://github.com/topepo/caret/issues/755 ".

library(alluvial)
plotCM <- function(cm){
  cmdf <- as.data.frame(cm[["table"]])
  cmdf[["color"]] <- ifelse(cmdf[[1]] == cmdf[[2]], "green", "red")  
  alluvial::alluvial(cmdf[,1:2]
                 , freq = cmdf$Freq
                 , col = cmdf[["color"]]
                 , alpha = 0.5
                 , hide  = cmdf$Freq == 0
 )
}

dd<-read.table(text="   GENE UNTREATED TREATED
1    G1      FOUR    FOUR
2    G2      FOUR    FOUR
3    G3      FOUR    FOUR
4    G4      FOUR   THREE
5    G5     THREE    NONE
6    G6     THREE    NONE
7    G7       TWO   THREE
8    G8       TWO     ONE
9    G9       ONE     ONE
10  G10       TWO     ONE", header=TRUE)

labels=unique(dd$UNTREATED)
d1 <- factor(dd$UNTREATED,labels=labels)
d2 <- factor(dd$TREATED,labels=labels)
confusionMatrix(d1,d2) %>% plotCM()

Hope it helps.

Thanks, Sam