Quick links: Documentation | Installation | R examples | Python examples

Showcase

Get all the goodies of UpSetR, but with full extensibility of ggplot2.

Example UpSet plot

Click here to display the source code

library(ggplot2)
library(ComplexUpset)

if(!require(ggplot2movies)) install.packages('ggplot2movies')
movies = ggplot2movies::movies
genres = c('Action', 'Animation', 'Comedy', 'Drama', 'Documentary', 'Romance')

upset(
    movies,
    genres,
    annotations = list(
        'Length'=ggplot(mapping=aes(x=intersection, y=length)) + geom_boxplot(),
        'Rating'=ggplot(mapping=aes(x=intersection, y=rating))
            # if you do not want to install ggbeeswarm, you can use geom_jitter
            + ggbeeswarm::geom_quasirandom(aes(color=log10(votes)))
            + geom_violin(width=1.1, alpha=0.5)
    ),
    queries=list(
        upset_query(
            intersect=c('Drama', 'Comedy'),
            color='red',
            fill='red',
            only_components=c('intersections_matrix', 'Intersection size')
        ),
        upset_query(
            set='Drama',
            fill='blue'
        ),
        upset_query(
            intersect=c('Romance', 'Drama'),
            fill='yellow',
            only_components=c('Length')
        )
    ),
    min_size=10,
    width_ratio=0.1
)

The full list of examples is available in the documentation; it also contains instructions for the use from Python.

Install

To get the most recent version, open R and run:

if(!require(devtools)) install.packages("devtools")
devtools::install_github("krassowski/complex-upset")

Alternatively, to get a stable CRAN release (which may be one version behind at times):

install.packages('ComplexUpset')

Or, if you use conda/mamba:

conda install -c conda-forge r-complexupset

How it compares to other packages?

  • UpSetR is a powerful tool and the pioneer in the UpSet visualisations; it was not designed to be extended with ggplot components. Unfortunately, the repository is no longer active (for two years now).
  • ggupset uses scales to convert a single plot to upsets plots, e.g. with scale_x_upset - a really nice approach, recommended for simple ggplot.
  • a cowplot-based upset was demonstrated in an online publication, however cowplot (a great tool at the time) got superseded by even greater tool: patchwork, which is what is used by ComplexUpset.
  • ComplexHeatmap also offers a way to generate UpSet plots with annotations; while not ggplot2-centered, it provides extensive customization options, with a clear API. It may be the best choice if you already use it for heatmaps.

Use whichever tool you find the most useful for your particular use. The rational of making this repository public is not only to share the code, but also to demonstrate how simple it is now to create complex visualisations with patchwork (without the need to learn the ggplot internals).

For the interactive use, check out the VCG/upset. Regardless of the tool chosen, you may want to cite (Lex et al, 2014) when using UpSet plots, especially in fields still dominated by Venn diagrams.

Testing

The unit tests are run with testhat, and the visual “doppelganger” tests use vdiffr.

The test cases for visual tests are auto-generated from the examples in the documentation; after changing or adding an example, please run:

  • scripts/prepare_for_tests.sh to generate updated test cases, and
  • scripts/manage_visual_tests.sh to validate the generated images.

Acknowledgments

Originally developed in course of a DPhil programme in Women’s & Reproductive Health at OxfordNuffieldWRH.