Help

How to do each task in DisCanVis, the terms the pages use, and worked examples.

Start here

Pick the task you came for.

DisCanVis puts germline and somatic variants on protein sequences, next to intrinsic disorder, motifs, domains, structure, predictors and cohort data, so that a variant is read in the context of the residue it hits. A change in a folded domain and one in a disordered linker are different events, and the pages show them apart.

Each task starts on one page and hands what you found, a protein, a residue, a variant or a set, to the next through the Next steps links at the end of an answer.

Not a diagnostic device. DisCanVis aggregates public data and predictors. Its automated classifications are evidence summaries; clinical decisions need your own protocols, laboratory validation and professional judgement.

Look up a protein

Type a name into the search box and the page it opens is the protein's Summary.

Type into the search box on Home, or the quick search in the menu bar of every other page. The box reads what you typed and takes you to the page that answers it; the line under the box says what it recognised (for example TP53 → TP53-201 · P04637).

  • A gene symbol, UniProt accession or entry name, Ensembl transcript or GENCODE name (TP53-201) opens that protein's Summary.
  • A variant (TP53 R175H, an rsID, a ClinVar VCV / RCV id, 7-140753336-A-T, an HGVS name) opens Variant interpretation.
  • A GO id or an exact GO or disease name opens Compare a set with that set.
  • Anything else goes to the Search table, which says what was tried and offers the nearest route when nothing matches.

The Summary page

/summary/<accession>/ is the per-protein report. Its header shows the protein's identifiers and three panels: Entry info (diseases, cancer types, curated sets, what DisCanVis holds for this isoform), UniProt info (read live from UniProt) and PPI network. A link to the older full-page Visual viewer is kept beside them. Four tabs:

  • Overview — the protein at a glance: the chain split into disordered and ordered stretches, whether variants fall in one more than the protein's own per-residue rate predicts, subcellular location, affected tissues, variants by source and context, and annotated regions. All · Somatic · Germline pathogenic change which variants are drawn. Next steps at the bottom carry the protein to Variant interpretation, AI Chat or the interaction network.
  • Explore — the main figure: every track on one residue axis, with the sequence and a 3D structure. See below.
  • Statistics — a dashboard of nine cards in which each count is set against a baseline: the protein's own residues in the same structural context, or the proteome. Origin (Germline / Somatic) and Scope (drag over the Overview figure to select a range) re-cut every card. It also holds the predictor diagnostics: how well each predictor separates germline pathogenic from benign variants, and how it agrees with the MAVE assays.
  • Tables — the record tables: Germline mutations, Somatic mutations and the matching Germline regions and Somatic regions, each with copy and CSV export. Each table builds when you open it.

Try it: the records behind a motif, as the Tables tab lists them.

Live

The records behind a motif

RAF1

The motif at RAF1 residues 254–263, as the Tables tab lists it: its name, its class and the residues it spans. The residue span is a filter you can remove.

Figure
RAF1: a real figure from the Summary page, built when you ask.
This view on the full page
/summary/RAF1-201/?tab=tables&table=elm
Open this in DisCanVis Back to RAF1

Explore

  • What stands out above the figure lists the regions and residues that differ from this protein's own background. Significantly mutated regions are drawn as one merged strip under the variants.
  • View chooses the figure: Summary, Structure, Variant effect, Orthologs or Custom. Each has one thin settings row of its own.
    • Structure: PDB | AlphaFold, a PDB entry picker, Mark entry (marks on the tracks the residues that entry resolves), Water & ligands, and chips for the ordered, disordered and "disordered, ordered in a complex" regions. A chip marks every such region at once. This is co-occurrence of a disorder call and a structure's span, not a computed interface.
    • Variant effect: Functional map | Compare predictors; Marks on map (Show, Germline, Somatic, Yours); a variant box and Open in VI.
    • Orthologs: taxonomic level and the species shown.
    • Custom: Regions (annotate a span, search a regex, PSSM or ELM motif, each region with its own colour and, optionally, its name on the track) and Copy link.
  • Settings: Simplified | Detailed (a few rows, or one row per annotation), Dense, Tracks (a drawer to switch tracks on and off, split or merge rows and reorder them), Presets (disorder-, variants- or motifs-focused, in Summary and Custom) and Export.
  • Export saves the figure as PNG or SVG, with the label column, legend and, in the Structure view, the 3D structure. Sizes include a square, 4:3 and journal single and double column widths in millimetres. Region marks can name the highlighted regions on the exported copy.
  • Copy link (in the Tracks drawer and in Custom) puts the view, its level, tracks, order and regions in a URL. ?position=175 or ?start=254&end=263 select and zoom to a residue or span.

Try Explore live

Each card opens a real figure when you ask for it. The numbered steps change what it shows; the box at the bottom runs the same view on your own protein.

Live

A hotspot in a folded domain, not in a disordered end

TP53

The Combined disorder bar is pink where TP53 is disordered and blue where it folds. Look at where the variants fall on it. The line above the figure sets the domain against the protein's own background.

Figure
TP53: a real figure from the Summary page, built when you ask.
This view on the full page
/summary/TP53-201/?view=summary
Open this in DisCanVis Back to TP53
Live

A motif in a disordered linker, hit by germline and somatic variants

RAF1

RAF1 is disordered between 196 and 343. Residues 254–263 carry germline and somatic variants and sit under motif annotations. The strip under the variants marks a significantly mutated region.

Figure
RAF1: a real figure from the Summary page, built when you ask.
This view on the full page
/summary/RAF1-201/?view=summary
Open this in DisCanVis Back to RAF1
Live

A disordered region that is resolved in a complex

CTNNB1

The N-terminus of β-catenin is disordered on its own. PDB entries resolve part of it inside a complex. This is a disorder call and a structure's span meeting, not a computed interface.

Figure
CTNNB1: a real figure from the Summary page, built when you ask.
This view on the full page
/summary/CTNNB1-201/?view=structure
Open this in DisCanVis Back to CTNNB1
Live

Do the predictors agree where the variants fall?

TP53

Variant effect scores every possible substitution: a bar per residue and, below, one cell per amino acid. Zoom in, then ask for one band per predictor and compare them under the variants.

Figure
TP53: a real figure from the Summary page, built when you ask.
This view on the full page
/summary/TP53-201/?view=pathogenicity
Open this in DisCanVis Back to TP53

The same Explore figure is embedded in the Protein context tab of Variant interpretation and in the Tracks view of AI Chat, so its controls are the same there.

Interpret a variant

One change, read at its residue, with the evidence kept apart from the call.

Variant interpretation reads one change at its residue: what covers it, what has been recorded or predicted there, and what the substitution does to the motifs around it. The page opens on one row: Protein, Variant, Interpret. The link /variant-interpretation?protein=TP53&variant=R175H fills it and runs it; every run writes its own call into the address bar so you can copy or bookmark it.

  • What you can type: a protein change (R175H, p.Arg175His), a stop-gain (R213*), a deletion, insertion, duplication or delins (K139_Q144del, K139dup), a frameshift (T155fs), or a genomic variant: an rsID, 7-140753336-A-T, a ClinVar VCV id, an HGVS c. or g. name. A protein is not needed for those.
  • Advanced (closed by default) holds pasted sequences, the indel and frameshift builders, the genome-build selector (Auto, GRCh38, GRCh37) and browsing a list of recorded variants by sample, variant id, gene, position or variant.
  • If the reference residue you typed is not the residue in the sequence, the page stops before interpreting and offers to re-run with the residue the sequence has.
Live

One variant on its protein map

BRAF

BRAF V600E: the residue is selected and the changed residue is drawn under the sequence. Variant interpretation reads the same variant and adds the evidence and the classification.

Figure
BRAF: a real figure from the Summary page, built when you ask.
This view on the full page
/summary/BRAF-220/?view=summary&variant=V600E
Open this in DisCanVis Interpret V600E Back to BRAF

The report

  • Overview: the variant on the protein map, two sentences built from the evidence that is present (where the residue sits, which evidence carries the call, what is missing), and the evidence ledger: one row per question, each with a direction (damaging, benign, uncertain, context, no data) that is never combined into a call. A count is shown against the protein's own per-residue rate.
  • Evidence: one card per source (Germline, Somatic, curated context, DisProt, PDB structure, disordered binding, phase separation, predictors, position profile, predicted sites, motifs) and a sortable table with CSV export. A source with nothing at the residue is listed under No data.
  • Classification: the automated ACMG/AMP criteria that the evidence meets, combined by the published rule. It is an evidence-based reading, not a clinical classification: the criteria do not read gene–disease validity, inheritance, segregation or functional data, and each criterion that would need them is listed as not evaluated. ClinGen's gene–disease validity and mode of inheritance are shown beside it as a gene-level fact (the ClinGen row of the evidence ledger, the ClinGen disease · inheritance column of Your data); a gene without one has not been curated by ClinGen, which is not the same as having no disease. What if… lets you add or re-weight a criterion and see the outcome.
  • Protein context: the Summary Explore figure with the residue marked. Sequence context: a window on the wild-type and mutant sequence, with copy and FASTA export. Download document saves the report.

Germline and somatic readings

There are two readings, shown behind a Germline | Somatic switch and never merged. The germline reading is the ACMG/AMP one. The somatic reading follows the ClinGen / CGC / VICC oncogenicity standard (Horak et al. 2022): it uses cancerhotspots.org (OS3, OM3, OP3) and CIViC (OS1, OM4) among its criteria and leads with a short cancer-evidence block: hotspot counts, recurrence against the protein's own rate, significantly mutated regions, the Census role, and the strongest CIViC evidence with its therapy, disease and PubMed id. A criterion that needs a curator, such as functional studies, is listed as not evaluated. One house rule departs from the standard and is named wherever it acts: a cancer hotspot that reaches OS3's counts and is clinically validated, either by a CIViC oncogenicity assertion on the very change (OS3 is then counted beside OS1) or by CIViC evidence at level A or B on it (a strong house criterion CV is then counted beside OS3), does not read as uncertain. Level C and D and functional evidence do not count. The switch opens on the reading that matches where the variant is recorded, so a known cancer hotspot opens on the somatic one.

Analyse your list

A list of proteins, regions or variants, read against each protein's own background.

Datasets › My sets reads a list of proteins, regions or variants: a table with a protein change, DisCanVis output, or a VCF or MAF with genomic coordinates in GRCh37 or GRCh38. Each card has Load an example.

  • The page finds the header, asks which columns hold the protein and the variant, and checks every row's reference residue against the sequence. Rows that cannot be placed stay in the tables with their reason.
  • The results are three tables, Variants, Regions and Proteins (plus Groups when the list names two or more groups), each set against the protein's own background. Variants can be read over the disordered ones alone, and synonymous or common (gnomAD frequency at or above 5 %) variants can be left out.
  • Interpret variants, or the ⓘ on a row, gives the same evidence and automated classifications as Variant interpretation; VI ↗ opens the row there, and Map beside the gene draws the list as a Your variants track in Summary.
  • What you upload is kept in this browser tab, not on the server. Close the tab and it is gone. A saved set can be picked in Compare a set like any other set.

Compare a set

A set of proteins, a GO term or a disease, with every count set against a background.

Datasets has two tabs: My sets (your sets, and the ready sets of the site) and Compare (a set against the proteome).

  • Simple is three choices and a button: Which proteins (the proteome, cancer drivers, DisProt, condensate and other curated sets, MAVE-measured proteins, or a GO term), Where in the protein (the whole protein, disordered regions, a kind of motif, domain, binding or phase-separating region, significantly mutated regions) and Which variants (all sources, Somatic, Germline pathogenic). Advanced adds diseases, several sets, thresholds and the motif screen.
  • The answer opens on the Table; each row is set against a background (a region's rate against the rest of its own protein, a protein's against the proteome) and says when there is none. A map above the table plots expected against observed; drag over it to keep only those rows. Figures is the second tab; a chart without a baseline is labelled Counts.
  • Your own proteins, regions and variants from Your data can be picked here as sets.

Search and browse

The catalogues: proteins, regions, diseases, somatic regions and the PPI network.

Search holds the catalogues. Each is a page of its own, linked along the top (Complexes and About too); Home and Summary reach them through the search box and Next steps.

  • Search proteins — a gene, accession or GO term gives the matching proteins with their disordered / ordered share, somatic and germline variant mix and the shares of missense, frameshift and indel variants per cancer source. Copy, CSV, Excel and print export; each row links to Summary, the PPI network and the REST answer.
  • Regions — every annotated region of one kind across the proteome, one row each, with variant counts, scores and structural context, filtered by a protein set, chromosome, GO term or disease. Live queries the database; Snapshot reads a pre-built table, much faster on large datasets and stable enough to cite (its build time is shown).
  • Germline diseases — OMIM and ClinVar disease tables.
  • Somatic regions — significantly mutated regions (iSiMPRe) by cancer cohort.
  • PPI network — a protein's interaction neighbourhood read through disease, cancer and structural evidence. Partners are picked by evidence. The structural shortlist reports that two proteins occur in the same PDB entry, not a computed interface; "no shared entries" and "no structural data" are different statements.

Most tables take column filters: numeric operators such as >100, and text matches. Check the filter row above the table.

About Search

What each view does and where its data come from.

Two ways in, one catalogue.

  • Find a protein — One gene, accession or GO term → the matching proteins with their disorder composition, mutation mix and cancer variant spectrum.
  • Browse regions — One annotation type → every region of it in the proteome, one row each, filterable by mutation source, score and structural context.

Live queries the database and returns the whole result for client-side sorting. Snapshot reads a pre-built table — much faster on the large datasets, and stable enough to cite. Snapshot built 2026-11-19 14:48 UTC.

Colours match Statistics and the per-gene Summary tracks.

Downloads & API

Bulk files, the same data by URL, and the proteome statistics.

  • Downloads — genome-wide TSV and FASTA tables by category and the mutation × region tables. PROVENANCE.json heads the files with the data release and the version of every source.
  • API — the same data by URL: /rest/… for one protein, region or position (positions are 1-based), the versioned /api/v1/analysis/… endpoints for Analysis, and the endpoints behind Summary's statistics and the Explore embed. It has clickable examples, a try-it panel and Python snippets.
  • The proteome statistics page draws disorder, somatic and germline variation, predictors and structure context over a saved catalogue, a GO term, a PPI neighbourhood or your own list.
  • Every JSON answer carries a meta.release block. Record the data release (shown in the footer) with any analysis.

Terms

The words the pages use, in one place.

Germline · SomaticThe origin of a variant. Germline is inherited: ClinVar and OMIM. Somatic arises in a tumour: TCGA, cBioPortal and the other cancer cohorts. A source's name is kept where a statement is about that source alone ("ClinVar pathogenic"). Disease names and cancer types keep their own words.
Record · variant · sampleA record is one row of one source (one ClinVar submission, one tumour sample's call). A variant is a distinct change. ClinVar numbers on Statistics count a variant once, however many submitters report it.
BaselineWhat a count is measured against: the protein's own per-residue rate in the same structural context, or the proteome. "39 records here" is an observation; whether it is more than this protein's background is the finding.
Disordered residue · disordered regionA residue is disordered when its combined disorder score is 2.0 or more. A region is disordered when more than 60 % of its residues are; that share is a house convention, not a published cutoff.
Badges C / P and G / M / O / DOn tracks and tables, C is curated or experimental evidence and P a computational prediction. The letter beside it is the level: G genome (variants), M protein-wide, O ordered context, D disordered context.
Published and convention cutoffsEach predictor is drawn against its own calibration. A cutoff marked published comes from a paper; one marked convention is a DisCanVis house choice. A dbNSFP rankscore is a percentile (more damaging than that share of the proteome), not a pathogenicity call, and is never read against a published raw-score cutoff.
— (a dash)No data. A value that is missing is never replaced by a default verdict.
Your dataWhat you upload to Analysis. It stays in your browser tab and is not stored on the server.

Worked examples

Real questions on real data. Each link runs the example.

What does TP53 R175H do? Variant · Variant interpretation
Which linear motifs sit inside a cancer hotspot? Set × regions · Compare a set
Where do my variants fall? Your data · Your data