UpSet plot in Julia with Makie

Unknown — Wed, 19 Mar 2025 00:00:00 +0000

UpSet plots</a> are a data visualization method that is used to show multiple intersecting sets. Conceptually, they provide similar information as a Venn diagram, but can be easier to comprehend when the number of sets is higher. Additionally, they provide the ability to show additional attributes of the sets. Here we'll implement a simple UpSet plot in Julia</a> using Makie</a>.</p>

We'll use CairoMakie, which can be used to render plots as SVG or PNG images. Additionally, we'll use the Combinatorics</a> package to get all possible combinations of the provided sets. First, make sure you have the necessary dependencies installed:</p>

using</span> Pkg
</span>Pkg.add("</span>CairoMakie</span>")
</span>Pkg.add("</span>Combinatorics</span>")
</span></code></pre>
Then we can add the packages to our namespace:</p>
using</span> CairoMakie
</span>using</span> Combinatorics
</span></code></pre>
The data for UpSet plots can come in various ways. Our function will use the following format as its main input:</p>
# The keys of the dictionary represent the combinations of
</span># the sets we want to visualize and the values are the number
</span># of elements common to all sets of the corresponding key.
</span>data = Dict(
</span>  ["</span>a</span>"] => </span>100</span>,
</span>  ["</span>b</span>"] => </span>93</span>,
</span>  ["</span>c</span>"] => </span>123</span>,
</span>  ["</span>a</span>", "</span>b</span>"] => </span>27</span>,
</span>  ["</span>a</span>", "</span>c</span>"] => </span>33</span>,
</span>  ["</span>b</span>", "</span>c</span>"] => </span>11</span>,
</span>  ["</span>a</span>", "</span>b</span>", "</span>c</span>"] => </span>3
</span>)
</span></code></pre>
For example, in the data above we have indicated that we have 100</em> elements in set a</strong> and 93</em> in set b</strong>, while the intersection between them contains 27</em> elements.</p>
We also will provide a vector of the base sets to the function, so that we can control the order in which they're shown:</p>
order = ["</span>a</span>", "</span>b</span>", "</span>c</span>"]
</span></code></pre>
Then, a basic UpSet plot can be accomplished with a function such as this one:</p>
function </span>upset</span>(data, order)
</span>  fig = Figure()
</span>
</span>  num_sets = length(order)
</span>
</span>  </span># top is the barplot part of the UpSet.
</span>  </span># bottom is the part that shows the set intersections.
</span>  </span># We reverse the given order, because we want the
</span>  </span># first given set to be at the top (highest y-value).
</span>  top = Axis(fig[</span>1</span>, </span>1</span>])
</span>  bottom = Axis(fig[</span>2</span>, </span>1</span>],
</span>                yticks = (</span>1</span>:num_sets, reverse(order)))
</span>
</span>  </span># We get all possible combinations, not just
</span>  </span># the ones that were given in the input data.
</span>  combs = collect(combinations(order))
</span>  num_combs = length(combs)
</span>  comb_counts = map(c -> get(data, c, </span>0</span>), combs)
</span>
</span>  set_index = Dict(zip(order, </span>1</span>:num_sets))
</span>
</span>  barplot!(top, </span>1</span>:num_combs, comb_counts, color=</span>:black</span>)
</span>
</span>  </span>for </span>(i, comb) </span>in</span> enumerate(combs)
</span>    </span># For each combination we get the row positions of the
</span>    </span># top and bottom sets. We use this information to plot
</span>    </span># lines that connect the dots that represent the sets.
</span>    min_y = min(map(s -> set_index[s], comb)...)
</span>    max_y = max(map(s -> set_index[s], comb)...)
</span>    lines!(bottom, [i, i], [num_sets - min_y + </span>1</span>, num_sets - max_y + </span>1</span>], color=</span>:gray</span>)
</span>    </span># We plot as large dots all the sets in the current combination.
</span>    </span>for</span> set </span>in</span> comb
</span>      scatter!(bottom, i, num_sets - set_index[set] + </span>1</span>, markersize=</span>15</span>, color=</span>:black</span>)
</span>    </span>end
</span>  </span>end
</span>
</span>  xlims!(top, </span>0.5</span>, num_combs+</span>0.5</span>)
</span>  xlims!(bottom, </span>0.5</span>, num_combs+</span>0.5</span>)
</span>
</span>  hidexdecorations!(top)
</span>  hidexdecorations!(bottom)
</span>
</span>  linkxaxes!(top, bottom)
</span>  hidespines!(top)
</span>  hidespines!(bottom)
</span>
</span>  </span># This controls the relative height of the two rows.
</span>  </span># This would make a fine additional parameter for our
</span>  </span># function so that we can easily tweak it, depending
</span>  </span># on the number of sets we want to plot.
</span>  rowsize!(fig.layout, </span>1</span>, Relative(</span>2</span>/</span>3</span>))
</span>  rowsize!(fig.layout, </span>2</span>, Relative(</span>1</span>/</span>3</span>))
</span>
</span>  fig
</span>end
</span></code></pre>
Now, we can run the function and obtain an UpSet plot:</p>
fig = upset(data, order)
</span>save("</span>upset_plot.png</span>", fig)
</span></code></pre>
Which produces the following plot:</p>
</p>