A colleague asked me this question: "I'm trying to find a way to find genes that overlap between three datasets. I have used intersect for two dataframes but can't seem to find a solution for three dataframes on google. Do you know any snazzy way of doing that?" I thought using the venn() function from the gplots package is a pretty snazzy solution, so that's what I recommended.

The venn() function is used to plot a Venn diagrams for up to 5 sets but you can use it to list intersections.

# install if necessary install.packages('gplots') library(gplots) a <- 1:10 b <- 3:12 c <- 5:14 d <- 7:16 e <- 9:18 my_list <- list(a, b, c, d, e) venn(my_list) # save object from venn my_venn <- venn(my_list, show.plot = FALSE) class(my_venn) [1] "venn" # which object class? library(pryr) otype(my_venn) [1] "S3" # what are the attributes names? names(attributes(my_venn)) [1] "dim" "dimnames" "intersections" "class" # use attr() to get intersections attr(x = my_venn, "intersections") $A [1] "1" "2" $E [1] "17" "18" $`A:B` [1] "3" "4" $`D:E` [1] "15" "16" $`A:B:C` [1] "5" "6" $`C:D:E` [1] "13" "14" $`A:B:C:D` [1] "7" "8" $`B:C:D:E` [1] "11" "12" $`A:B:C:D:E` [1] "9" "10" # get specific intersections attr(x = my_venn, "intersections")$`A:B` [1] "3" "4"

*A bit hard to visualise the intersections when there are five lists*.

Notice that **attr(x = my_venn, "intersections")$ A:B** gives you only 3 and 4 and not 3, 4, 5, 6, 7, 8, 9, and 10? This is because 3 and 4 are the only intersections that are unique to A and B, when taking the other intersections into account.

intersect(a, b) [1] 3 4 5 6 7 8 9 10 # an element in the union of C, D, and E may or may not intersect A and B # however, if they do intersect any element in A and B, then that element is not # unique to intersect(a, b) setdiff(intersect(a, b), union(union(c, d), e)) [1] 3 4 # and finally the non-snazzy way of intersecting three lists intersect(a, intersect(b,c)) [1] 5 6 7 8 9 10

This work is licensed under a Creative Commons

Attribution 4.0 International License.

Hi Dave,

yesterday a new preprint exactly for this task plus plotting came out:

http://biorxiv.org/content/early/2017/02/18/109728

Might be of interest.

Cheers,

Andreas

Hi Andreas,

thanks for that. The plots look great and it even comes as a Shiny app!

Cheers,

Dave

I think you can use jvenn, it's very simple and useful!

http://bioinfo.genotoul.fr/jvenn

Thanks, it does look like a nice tool.