Compare a set

Ask where the variants of a set of proteins land (ordered or disordered, inside which annotated regions) and whether that is more than the proteome would give.

Narrows the proteins above to those annotated with the term.
Load an example

A disease, several sets, thresholds or a motif screen?

Human proteome

Settings changed — open any step above and click Apply & show results.
—
proteins
—
disordered residues, mean
—
missense / residue, mean

Proteins
—
Regions
—
Studies
—
Mutations
—
Counting
Rows
Regions
Columns
How this table is computed
One row per variant (or per file row, with One per variant off), with where it sits and what is already known about it. Disordered only keeps the variants at disordered residues (the combined disorder call). ⓘ reads one variant here; VI ↗ opens it in Variant Interpretation; a protein's link opens Summary with your whole list as the Your variants track.
One row per annotated span on the proteins your list touches. Expected is how many of that protein's uploaded variants would fall in the span if they were spread evenly along the chain; O/E is the ratio, tested one-sided (binomial) and corrected across the spans tested (Benjamini–Hochberg). A span is marked enriched at q < 0.05 and only on a protein with at least 3 variants. A variant is counted once however many rows of the file carry it — Rows says how many. Synonymous changes are not counted. Spans are those DisCanVis summarises — a span is listed once any database has a variant in it. Evidence says what kind of statement the annotation is: curated, predicted, curated + predicted or derived. With Disordered only the whole rollup is redone over the disordered variants, with the baseline narrowed to the protein's disordered residues.
One row per protein. Expected in disorder is the protein's placed variants times the share of its residues called disordered; O/E and q as for regions, each variant counted once and synonymous changes not counted. Open a row (▸) for every span on the protein, including the ones no uploaded variant hits. With Disordered only the counts are of those variants, while the in disorder / ordered split and its test stay over the whole list.
One row per protein and region: the variants a PDB entry resolves together, the entries that resolve them and the other proteins in those entries: a co-occurrence in an entry, not a computed interface. ◆ counts the entries where the residue is disordered and resolved, with a partner protein. Expand a row for its entries; Open structure opens the entry that resolves the most variants in the complex view with your data marked.
Each group of your list, then two of them compared the way cohort mutation studies do (maftools mafCompare): for every protein, variant or annotated span, carriers against non-carriers per group, Fisher's exact test (two-sided), the odds ratio with a 95% interval, and q across the features tested (Benjamini–Hochberg). A feature is tested only when one group has at least the minimum number of carriers. The carrier is the sample when a sample column was mapped, otherwise the variant. Synonymous changes are not counted.

Your variants, observed against expected

Reading your list…

Filtered by the figure: ·

Columns
API
Docs

            
Running…