1
$\begingroup$

As i have multiple comparison so I m trying this UpSet plot, but it needs data to be formatted in particular way, one of the ways is as a list, so now the issue is since between one comparison the no of genes coming out may not be same as the other one, so there is a difference in dimension so if i bind them as data frame then i get dimension error, so the best way to get through is make list so what i have tried so far

A <- read.csv("GMP_MONO_UP_DOWN_CDS_EPI.txt",header = TRUE)
B <- read.csv("CMP_GMP_UP_DOWN_EPI.txt",header = TRUE)
C <- read.csv("HSC_BLAST_UPDOWN_EPI.txt",header = TRUE,sep = '\t')

dim(A)
dim(B)
ONE <- A[c(1)]#############just taking out the gene column########
TWO <- B[c(1)]#############just taking out the gene column########
THREE <- B[c(1)]#############just taking out the gene column########
dim(ONE)
dim(TWO)
b1 <- apply(ONE,2,list) 
b2 <- apply(TWO,2,list)

Now this is the code from their help page

upset(movies, nsets = 7, nintersects = 30, mb.ratio = c(0.5, 0.5),
      order.by = c("freq", "degree"), decreasing = c(TRUE,FALSE))

so the data "movies" class is data frame, there is not dimension issue as such as in my case if i want to intersect two comparison which may not have the same length, so how do i overcome the issue any suggestion or help would be highly appreciated. Update my question with one more data to add.

$\endgroup$
3
$\begingroup$

Don't you just need a binary matrix with presence/absence data for all of the genes?

library(dplyr)
combined <- full_join(data.frame(gene=ONE, GMP_MONO=1),
                      data.frame(gene=TWO, CMP=1))
combined[is.na(combined)] <- 0

Edit: To loop over a large number of data frames, the easiest solution is to use the reduce function from the purrr package:

combined <- reduce(list(data.frame(gene=c('gene1', 'gene2', 'gene3'), set1=1),
                        data.frame(gene=c('gene3', 'gene4', 'gene5'), set2=1),
                        data.frame(gene=c('gene1', 'gene4'), set3=1)
                       ), full_join)
combined[is.na(combined)] <- 0
combined
#    gene set1 set2 set3
# 1 gene1    1    0    1
# 2 gene2    1    0    0
# 3 gene3    1    1    0
# 4 gene4    0    1    1
# 5 gene5    0    1    0
$\endgroup$
4
  • $\begingroup$ Im getting NA in the CMP column in my combined data-frame after doing full_join , im getting 0 as well but not sure why NA.. $\endgroup$
    – kcm
    May 31 '18 at 5:19
  • $\begingroup$ Oh yeah. I forgot about that. See updated answer $\endgroup$ May 31 '18 at 5:51
  • $\begingroup$ yeah now its working but i do see error when i try to do more than ,which is apparently wont happen with full_join ,lets say i do want to intersect , more than three data frame and make a binary matrix can you suggest what more to add to your code? $\endgroup$
    – kcm
    May 31 '18 at 6:40
  • $\begingroup$ see updated answer $\endgroup$ May 31 '18 at 7:33
7
$\begingroup$

We can plot using lists as well, using fromList function, see example:

library(UpSetR)

myGeneSets <- list(
  set1 = c("gene1","gene2","gene3"),
  set2 = c("gene1","gene4","gene5", "gene7"),
  set3 = c("gene1","gene7")
  )

upset(fromList(myGeneSets))

fromList

A function to convert a list of named vectors to a data frame compatible with UpSetR.

fromList(myGeneSets)
#   set1 set2 set3
# 1    1    1    1
# 2    1    0    0
# 3    1    0    0
# 4    0    1    0
# 5    0    1    0
# 6    0    1    1
$\endgroup$
1
  • 1
    $\begingroup$ another way to do the stuff cool ..i will use this as well $\endgroup$
    – kcm
    Jun 4 '18 at 17:05
1
$\begingroup$

You need to set a common number of rows, one for each gene (make sure that the gene identifiers are the same for both files). Then in the first column you can set 1 for the file A, and 0 otherwise. In the second column you can do the same. Then you add this dataframe as the input of the upset function.

genes <- unique(c(ONE, TWO))
df <- data.frame(A = ifelse(genes %in% A, 1, 0), B = ifelse(genes %in% b, 1, 0))
upset(df, ...)
$\endgroup$
5
  • $\begingroup$ Your solution does work but it looks like what ever intersection it does it gives only one common element ..im not sure what is wrong.. $\endgroup$
    – kcm
    May 31 '18 at 6:56
  • $\begingroup$ Why would you expect more than one element? Could you share the data of ONE and TWO? Otherwise could you do length(intersect(ONE, TWO))? $\endgroup$
    – llrs
    May 31 '18 at 7:02
  • $\begingroup$ as I have multiple comparison between lineage type , the ONE ,TWO and now which i added later THREE they are just vector containing my gene list ,so im using only gene list .. $\endgroup$
    – kcm
    May 31 '18 at 7:14
  • $\begingroup$ Then you need to add a third column and genes should be made with genes <- unique(c(ONE, TWO, THREE)), could you post your data $\endgroup$
    – llrs
    May 31 '18 at 7:22
  • $\begingroup$ I will post the data ,update the question with my data. $\endgroup$
    – kcm
    May 31 '18 at 10:02

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.