Accept binary matrix input or create helper function for conversion

[moldach]

A number of tools, like SURVIVOR, create binary matrix. Tools like ComplexHeatmap's UpSet plot can either accept input from A) lists of sets (like upsetjs_r) or B) binary matrix.

It would be nice if upsetjs_r could either accept input from binary matrices or contain a helper function to convert binary matrix data.

I don't believe ComplexHeatmap (or any other package I know of) has a function to do this at the moment: https://support.bioconductor.org/p/131616/

[sgratzl]

can you provide a sample dataset, please?

[moldach]

Sure:

library(tibble)
mat <- tribble(
  ~set1, ~set2, ~set3,
   1,   1,   0,
   0,   0,   1,
   0,   1,   1,
   0,   0,   1,
   0,   0,   1,
   0,   1,   1,
   1,   0,   1,
   0,   1,   1,
   0,   0,   1,
   0,   0,   1,
   1,   1,   1,
   1,   0,   0,
   0,   0,   1,
   0,   1,   0,
   1,   1,   1,
   0,   1,   0,
   0,   1,   1,
   0,   1,   0,
   0,   0,   1,
   0,   0,   1
)

So either accept that type of data or a helper function to convert it into a format that upsetjs accepts, e.g.:

set.seed(123)
listInput = list(a = sample(letters, 5),
          b = sample(letters, 10),
          c = sample(letters, 15))
upsetjs() %>% fromList(listInput) %>% interactiveChart()

[sgratzl]

https://upset.js.org/integrations/r/articles/basic.html#data-frame-input doesn't work?

[moldach]

Sorry I never saw that documentation - I still cannot find a link for it on the README on the github repo.

One more thing, I've noticed the following discrepancy.

library(tibble)
t <- tribble(
  ~set1, ~set2, ~set3,
   1,   1,   0,
   0,   0,   1,
   0,   1,   1,
   0,   0,   1,
   0,   0,   1,
   0,   1,   1,
   1,   0,   1,
   0,   1,   1,
   0,   0,   1,
   0,   0,   1,
   1,   1,   1,
   1,   0,   0,
   0,   0,   1,
   0,   1,   0,
   1,   1,   1,
   0,   1,   0,
   0,   1,   1,
   0,   1,   0,
   0,   0,   1,
   0,   0,   1
)

Both VennDiagram and ComplexHeatmap produce the same result but upsetjs was producing a different image.

VennDiagram

library(VennDiagram)
venn.diagram(list(Set1=which(t[,1]==1), 
                  Set2=which(t[,2]==1), 
                  Set3=which(t[,3]==1)), 
             fill = c("#DDAA33", "#BB5566" ,"#004488"), 
             alpha = c(0.5, 0.5, 0.5), cex = 2, lty =2, 
             filename = "VennDiagram.tiff")

ComplexHeatmap

m = make_comb_mat(t)
ss = set_size(m)
UpSet(m, set_order = order(ss), comb_order = order(-comb_size(m)))

UpSetJS

upsetjs() %>% fromDataFrame(t)

According to ComplexHeatmap's documentation:

intersect mode: 1 means in that set and 0 is not taken into account, then, 1 1 0 means a set of elements in set A and B, and they can also in C or not in C (intersect(A, B)). Under this mode, the seven combination sets can overlap.

so,

m = make_comb_mat(t, mode="intersect")
ss = set_size(m)
UpSet(m, 
           set_order = order(ss),
           #comb_order = order(comb_degree(m), -cs),
           comb_order = order(-comb_size(m)))

So for ComplexHeatmap the default is distinct

distinct mode: 1 means in that set and 0 means not in that set, then 1 1 0 means a set of elements both in set A and B, while not in C (setdiff(intersect(A, B), C)). Under this mode, the seven combination sets are the seven partitions in the Venn diagram and they are mutually exclusive.

How can you mimic the distinct output produced by VennDiagram and ComplexHeatmap with UpSetJS?

[sgratzl]

so to understand it correctly

m = make_comb_mat(t) ss = set_size(m) UpSet(m, set_order = order(ss), comb_order = order(-comb_size(m)))

if you label the vertical bars in this chart it would read the following:

• setdiff(set3, union(set1, set2))
• setdiff(intersect(set3,set3), set1)
• setdiff(set2, union(set1, set3))
• intersect(set3, set2, set3)
• setdiff(intersect(set2, set1), set3)
• setdiff(intersect(set3, set1), set2)
• setdiff(set1, union(set3,set2))

right?

[jokergoo]

E.g., in ComplexHeatmap, for a pattern

A B C
x x -

the "distinct" mode corresponds to setdiff(intersect(A, B), C) and the "intersect" mode corresponds to intersect(A, B),

and similar for the pattern:

A B C
x - -

the "distinct" mode corresponds to setdiff(setdiff(A, B), C), or setdiff(A, union(B, C)), the "intersect" mode corresponds to intersect(A) which is A.

You can find the graphical explanations here:

[moldach]

Thank you for clarification @jokergoo

[sgratzl]

thx. I just implemented a generateDistinctIntersections() function.

can you verify that https://upset.js.org/next/integrations/r/articles/combinationModes.html is now showing the values you would expect in the different modes?

[moldach]

That's awesome @sgratzl , thanks for being so quick with this enhancement!

The function may not have been included in the most recent build though.

I deleted the folder from R/win-library/4.0 and tried to reinstall via: devtools::install_url("https://github.com/upsetjs/upsetjs_r/releases/latest/download/upsetjs.tar.gz")

but I get an error:

t <- read.csv("data/survivor_comparison_matrix.csv")
library(upsetjs)
upsetjs() %>% 
  fromDataFrame(t) %>% 
  generateDistinctIntersections() %>%   
  interactiveChart()

Error in generateDistinctIntersections(.) : could not find function "generateDistinctIntersections"

[sgratzl]

yeah it is not released yet but in the develop branch. You can download a current build e.g. from https://github.com/upsetjs/upsetjs.github.io/raw/master/next/integrations/r/upsetjs.tar.gz

[sgratzl]

version 1.1.0 has been released containing the generateDistinctIntersections function