UpSet Plot Generator for Overlapping Sets

What is an UpSet plot?

An UpSet plot shows how several sets overlap. Instead of overlapping circles, it lists each intersection as a column: a dot matrix underneath marks which sets the intersection contains, and a bar on top shows how many elements fall into it. Horizontal bars beside the matrix give the size of each set. Because every intersection gets its own column, an UpSet plot stays readable for five, ten, or more sets, where a Venn diagram has to draw 31 regions for five sets and 63 for six. Alexander Lex, Nils Gehlenborg and colleagues introduced the technique in 2014, and it is now standard for comparing gene lists, variant callers, peak sets, database contents, and patient conditions.

How to read an UpSet plot

• Each column is one intersection. The filled dots in that column, joined by a vertical line, are the sets it includes; gray dots are sets it excludes.
• The bar above the column is the intersection size. By default it counts elements in exactly those sets and no others, the same number a Venn diagram prints in the matching region.
• A column with a single dot holds the elements unique to that set.
• The horizontal bars on the left are the set sizes: all elements in a set, summed over every intersection that contains it.
• Columns are usually sorted by size, so the largest overlaps come first. Sorting by degree groups single sets, pairs, triples, and so on.
• Intersections with no elements are not drawn, so the plot only shows the overlaps that exist.

UpSet plot vs Venn diagram

A Venn diagram is the most intuitive picture for two or three sets: the regions map directly onto the overlaps, and readers need no explanation. With four sets the circles must become ellipses to show all 15 regions, and with five or more the regions get too small to label or compare. An UpSet plot shows the same numbers as bars that share one baseline, so sizes are easy to compare and to sort, and any number of sets fits. A good rule: use a Venn diagram for up to three sets, and an UpSet plot from four sets on, or whenever the exact sizes matter. If you already have Venn region counts, enter them as intersection names with counts and the generator turns them into an UpSet plot.

Exclusive and inclusive intersection sizes

• Exclusive (distinct) size: elements in exactly these sets and in none of the others. This is the default here, in UpSetR, and in the ComplexHeatmap "distinct" mode, and it matches Venn diagram regions. Exclusive sizes add up to the total number of elements.
• Inclusive size: elements in at least these sets, whatever else they belong to (the ComplexHeatmap "intersect" mode). The inclusive size of A and B counts everything in both A and B, including elements also in C.
• Inclusive sizes overlap, so they do not add up to the total, and the bar for a single set equals its set size.
• Write "inclusive intersections" in your instructions to switch modes. The generator computes both and states which one is drawn.

What does this UpSet plot generator calculate?

The generator reads your file in one of three formats, removes duplicate element-set pairs, and builds the membership of every element. It counts each set, every observed combination of sets, and both the exclusive and inclusive size of each intersection. It then sorts the intersections (by size by default, or by degree or set order), keeps the top 30 (up to 60 on request, or only those above a minimum size), and reports how many intersections and elements were left out, so the plot never hides data silently. In membership-matrix files it also counts elements that belong to no set. The image model draws exactly these numbers. It does not test whether an overlap is larger than expected by chance and adds no p-values.

Three ways to format UpSet plot data

• Membership matrix: one row per element, an optional ID column, and one column per set holding 1/0, TRUE/FALSE, or yes/no. Every such column is treated as a set. This is the format of UpSetR data frames and of from_indicators in the Python UpSetPlot package.
• Element and set pairs (long format): one row per element and set, for example "PT001, Hypertension" and "PT001, Type 2 diabetes". Map the element column and the set column. This matches gene lists stacked into one table, like UpSetR’s fromList or UpSetPlot’s from_contents.
• Intersection counts: one row per intersection with names such as "A&B&C" (also "A ∩ B", "A|B", "A;B") and an exclusive count. Map the intersection column and the count column. This is UpSetR’s fromExpression format, and the format to use for Venn region counts.
• Up to 20 sets and 1,200 rows are read. Sets beyond 20 are listed in a warning.

Where UpSet plots are used

• Differentially expressed genes shared across treatments, time points, or tissues, typically from DESeq2, edgeR, or limma results.
• Variants or peaks called by several tools or found in several samples.
• Proteins or metabolites identified across replicate runs or extraction methods.
• Comorbidities, symptoms, or adverse events that co-occur in the same patients.
• Species or taxa shared across sites, seasons, or hosts.