About & cite

How to cite DisCanVis, which data release you are reading, what changed recently, and who supports it.

You are on DisCanVis 2.0; the previous release stays available as Version 1.

How to cite

Use the references below for the portal and for related methods papers. Format: full title, authors, journal, date, DOI.

Also give the data release (2026.11, shown in the footer), the versions of the sources you rely on (Data sources) and the date you accessed the site.

DisCanVis: Integrating structural and functional annotations for the better understanding of the effect of cancer mutations located within disordered proteins

Norbert Deutsch, Mátyás Pajkos, Gábor Erdős, Zsuzsanna Dosztányi

Department of Biochemistry, Institute of Biology, ELTE Eötvös Loránd University, Budapest, Hungary · zsuzsanna.dosztanyi@ttk.elte.hu

Protein Science, 30 November 2022.
DOI: 10.1002/pro.4522

Pathogenic variations illuminate functional constraints in intrinsically disordered proteins

N. Deutsch, G. Erdős, Zsuzsanna Dosztányi

iScience 29 (4), 115215 (2026).
DOI: 10.1016/j.isci.2026.115215

AIUPred–Binding: Energy Embedding to Identify Disordered Binding Regions

G. Erdős, N. Deutsch, Zsuzsanna Dosztányi

Journal of Molecular Biology, 1 August 2025.
DOI: 10.1016/j.jmb.2025.169071

Data sources & release

Data release 2026.11
Data release
2026.11
heads every download and REST answer (meta.release)
Assembled
2026-11-16
portal version 2.0
Genome build
GRCh38, GENCODE v44
the one coordinate system
Sources
44
in 8 groups, below

Source date is the provider's release date where the pipeline recorded it, otherwise the date of the pipeline file (file YYYY-MM-DD). Loaded is when the table was last (re)loaded in the DisCanVis database. unknown means the value was not recorded: it is not a guess, and a later reload will fill it in. Hover a version for the file or log it was read from; hover a note for the full text. Cite the data release together with the source versions you rely on.

Somatic variants

8 sources
Versionunknown
Sourcefile 2026-09-14
Loadedunknown
A GRCh38 MAF mapped onto GENCODE v44; the GDC data release was not recorded.
Versionunknown
Sourceunknown
Loadedunknown
The COSMIC release was not recorded.
Versionunknown
Sourcefile 2026-09-14
Loadedunknown
Studies lifted from GRCh37 to GRCh38 by the pipeline (rows that cannot be placed are dropped, never guessed). The cBioPortal data version was not recorded.
Version24Q4
Sourcefile 2026-06-30
Loadedunknown
Model table 24Q4. Stored as database='DepMap' beside the other mutation sources.
Versionunknown
Sourcefile 2026-06-30
Loadedunknown
Census and Compendium role tables kept in the pipeline repository.
Significantly mutated regions derived
Versionderived from the somatic sources
Source—
Loaded2026-11-16
Computed from the mutation tables above and stored in the browse snapshot, rebuilt 2026-11-16.
Versionv2
Sourceunknown
Loaded2026-11-16
Genes with no main isoform (e.g. H3F3A, HIST1H3B) and residues whose reference letter differs from the isoform are not loaded.
Versionnightly 2026-11-16
Source2026-11-16
Loaded2026-11-16
CC0 (Griffith et al. 2017). Single-residue missense changes and residue-level profiles on main isoforms only; fusions, combined profiles, indels, truncations and gene-level profiles are not loaded, nor are rows CIViC flags. Its Oncogenic / Likely Oncogenic assertions are the previously classified changes the somatic reading's OS1 and OM4 read; clinical evidence is context, not points. The nightly was used because the 01-Nov-2026 monthly release holds 35 of the 50 oncogenic assertions.

Germline & clinical variants

3 sources
VersionVCF file of 2026-03-09
Source2026-03-09
Loadedunknown
Release is the date of the ClinVar VCF the pipeline mapped. Submission dates come from the submission summary (joined by VariationID). The database load date is not recorded (the load-stage log was last updated 2026-04-04); the ClinVarDisease table may still hold an earlier load.
Versionhumsavar 2026_02
Source2026-06-10
Loadedunknown
Gene-disease summary per protein. The per-variant OMIM table is empty in the database, so there is no OMIM variant track.
Versionexport of 2026-11-17
Source2026-11-17
Loaded2026-11-17
Gene-level: one row per curated gene–disease pair with its mode of inheritance, shown beside OMIM on the Variant Interpretation ledger and in Analysis › Your data. 3,027 of the curated genes match a main isoform by gene symbol; 53 do not (mitochondrial tRNA / rRNA and microRNA genes, and symbols this database holds under another name) and are not shown. A gene with no row has not been curated by ClinGen.

Population variants

2 sources
Version4.1 (joint)
Sourcefile 2026-09-15
Loadedunknown
Allele frequency and popmax AF read from the dbNSFP 5.3.1a export, not from a separate gnomAD file. dbSNP rsIDs: dbSNP 155 (dbSnp155Common.bigBed, 2026-06-17).
Version155 (Common)
Sourcefile 2026-09-14
Loadedunknown
Polymorphism track (rsID and allele frequency).

Disorder & phase separation

6 sources
Versionunknown
Sourcefile 2026-07-09
Loadedunknown
Fetched from the MobiDB API; the MobiDB release was not recorded.
Versionunknown
Sourcefile 2026-07-08
Loadedunknown
Fetched from the DisProt API; the DisProt release was not recorded.
VersionIUPred3, AIUPred, AIUPred-Binding, ANCHOR2
Sourcefile 2026-07-01
Loadedunknown
Computed by the DisCanVisFlow pipeline on the GENCODE v44 sequences. Tool versions are not recorded beyond the names.
Combined disorder derived
Version5-tier rule of 2026-07-27
Sourcefile 2026-07-27
Loadedunknown
MobiDB + RSA + IUPred + Pfam in five tiers. As recorded on 2026-11-05 the legacy web table (2024-12-10 rule) was not yet replaced by the pipeline's 2026-07-27 vector; the per-residue profile used by the Analysis maps is read from the newer one. Which rule a given page reads is stated on that page.
VersionDeepCoil, probability > 0.5
Sourcefile 2026-07-01
Loaded2026-11-16
CoiledCoils table reloaded 1-based on 2026-11-16 (the earlier load was the same predictor, 0-based). DeepCoil stays below 0.5 across some long coiled coils (MYH7 gets none).
VersionPhaSepDB v2.1; LLPSDB unknown
Sourcefile 2026-07-20
Loadedunknown
Auto-fetched and normalised by the pipeline. PhaSePro regions are listed under MFIB, DIBS and PhaSePro. The LLPSDB release was not recorded.

Motifs, PTMs & binding

6 sources
Versioninstances 2023, classes 2025, switches 2023
Sourcefile 2026-09-22
Loadedunknown
Version is the year in the file name; the exact ELM release dates were not recorded. The class catalogue (regex, GO compartments) was loaded for the motif scan on 2026-10-08 (manage.py load_elm_class_info). ELM data are non-commercial.
Versionpredicted_elm_dataset (iScience 2026 supplement)
Source2026
Loadedunknown
Supplementary table of Deutsch, Erdos, Dosztanyi, iScience 2026. The PEM_Core_Motifs table holds duplicate rows in the browse snapshot (known, guarded in Analysis).
Versionrelease of 2026-10-09
Source2026-10-09
Loaded2026-11-03
A MoMAP release is a snapshot; the table is replaced as a whole.
ScanSite computed
VersionScanSite 4.0
Sourcefile 2026-07-01
Loadedunknown
Kinase and binding-site predictions, not observed sites.
Versionunknown
Sourcefile 2026-09-22
Loadedunknown
Parsed copies kept in the pipeline repository; the database releases they were taken from were not recorded. PhaSePro: https://phasepro.elte.hu/.
Versionunknown
Sourcefile 2026-09-22
Loadedunknown
Licensed source files kept outside git; the releases were not recorded.

Structure & interactions

4 sources
Versionunknown
Sourcefile 2026-09-22
Loadedunknown
Fetched from the InterPro REST API; the InterPro / Pfam release was not recorded.
VersionSIFTS bulk file of 2026-06-17
Sourcefile 2026-06-30
Loaded2026-11-16
PDBRegional was reloaded 1-based from the pipeline's SIFTS spans on 2026-11-16. PDBStructure is an older, 0-based load (shifted +1 where it is read). Structural co-occurrence in a PDB entry is not a computed interface.
VersionAlphaFold DB v6
Sourcefile 2026-07-01
Loadedunknown
Per-residue pLDDT of the human proteome archive, version 6.
Versionunknown
Sourcefile 2026-10-20
Loadedunknown
In practice HIPPIE provides the coverage; BioGRID and IntAct archives parse almost no human pairs (open pipeline defect). The source release dates were not recorded.

Predictors & conservation

8 sources
Version5.3.1a
Sourcefile 2026-09-15
Loadedunknown
GRCh38 build, 40 rankscores loaded into DbnsfpRankscore; raw scores read on demand from the indexed file. CC BY-NC-ND 4.0: every predictor is gated by its own licence flag.
Versionunknown
Sourcefile 2026-09-22
Loadedunknown
The hg38 isoform file of 2024-04-18 mapped onto GENCODE transcripts. The standalone table is incomplete (316 main isoforms without rows, 2 696 partial; checked 2026-10-02).
Pathogenicity per residue (legacy table)
Versionunknown
Sourceunknown
Loadedunknown
An older table of predictor scores per residue, being replaced by dbNSFP rankscores. The release of its inputs was not recorded.
Versionunknown
Sourcefile 2026-06-30
Loadedunknown
Score sets are not comparable across assays; scores are rescaled within each set. Licences checked 2026-08-24.
Versionunknown
Sourcefile 2026-06-30
Loadedunknown
A separate aggregate with its own terms; the ProteinGym version was not recorded.
ACMG/AMP automated reading derived
Versioncomputed per request
Source—
Loaded—
Computed from the other sources on this page with the cutoffs in calibration.py (published or house convention, stated per cutoff); not a stored dataset.
Versionunknown
Sourcefile 2026-07-01
Loadedunknown
Genome-wide phastCons bigWig mapped onto proteins. The loaded track saturates (42 % of TP53 residues at the ceiling).
GOPHER conservation (taxonomic levels) computed
Versionunknown
Sourcefile 2026-07-01
Loadedunknown
A pre-computed multi-level table supplied to the pipeline.

Function, ontologies & reference

7 sources
Versiongo.obo releases/2026-05-19
Source2026-05-19
Loadedunknown
GOA human annotations generated 2026-05-21 on the 2026-05-19 ontology.
Version2026_02
Source2026-06-10
Loadedunknown
Release 2026_02 feeds regions of interest, binding sites and the variant index. The reference FASTA files that map GENCODE transcripts to UniProt entries are older (2023-07-27; curated isoforms 2024-03-21) and their release was not recorded. Entry text on Summary and the Variant Interpretation header is read live from the UniProt REST API.
Versionreleases/2025-02-04
Source2025-02-04
Loadedunknown
Groups ClinVar disease names into organ-system categories (DiseaseOntology table, 26 144 terms). The grouping rule is a DisCanVis convention.
VersionGTEx v8 (2017-06-05)
Source2017-06-05
Loaded2026-11-18
Gene-level: one row per GTEx gene with its median TPM in each of 54 tissues. A tissue counts as expressed at a median of 1 TPM or more, a common convention and not GTEx's own cutoff. Joined to proteins by gene symbol.
Versionlive
Source—
Loaded—
Read at question time and cached per record (LiteratureRecord). Text-mined, never proof. Queries send the gene or variant name to NCBI and EBI.
Versionv44
Source2023-03-01
Loadedunknown
Ensembl 110, GRCh38. The server's one coordinate system and the source of every Protein_ID (transcript name, e.g. TP53-201). Date of the load into the database not recorded.
Genome map (GRCh37 / GRCh38) derived
Versionbuilt on GENCODE v44; hg19ToHg38 UCSC chain
Sourcefile 2026-09-22
Loaded2026-11-03
Places genomic and identifier input (VCF, MAF, HGVS, rsID, ClinVar ids) on proteins. No runtime liftover.

Database scale

Current deployment totals (canonical proteome; refreshed with the deployed data build). Somatic = merged cohort variants (TCGA, cBioPortal and the other cancer sources); Germline = ClinVar variant rows; OMIM = same roll-up as the Germline diseases table.

Main isoforms
19,837
Proteome residues
11,191,792
Somatic variant rows
2,503,165
Merged cohort tables (missense, frameshift, indel); counts are records, not distinct variants
Germline variant rows (ClinVar)
2,941,032
One row per record with its significance and disease; a variant can have several rows
Pathogenicity predictors
15
Shown here, of 41 dbNSFP rankscores loaded — the rest are held back by their licence
MAVE assays
679
Measured variant effects from MaveDB, mapped onto these transcripts
Curated disordered regions
5,644
DisProt, plus 2175 phase-separating regions from LLPSDB
Significantly mutated regions
4,667
Spans a density test flagged in the somatic cohorts
OMIM disease rows
4,605
Gene–disease rows in the Germline diseases table

Figure credits

Species silhouettes (Summary › Orthologs) are from PhyloPic, downloaded once and served from this site. Only CC0, public-domain and CC BY images are used: 33 are CC0 / public domain (no credit required; the authors are listed in the manifest with each image's id and licence), and the Bos taurus silhouette (drawn from Bos primigenius) is by Ivan Iofrida, licensed CC BY 4.0. A species with no usable image is shown as a text label.

Release notes & recent changes

Portal v2.0

Concise, user-facing changelog. For reproducible analyses, record the data release shown in the footer and in the meta.release block of API results, the source versions listed under Data sources, the access date and the exact query/filter settings.

2026-11-16 · data release

Rebuilt in this release · 5 tables

  • PDBRegional (PDB spans) (2026-11-16): Reloaded 1-based from the pipeline's SIFTS spans; backed up first.
  • Coiled coils (CoiledCoils table) (2026-11-16): Reloaded from the DeepCoil run, 1-based.
  • cancerhotspots.org v2 (hotspot residues) (2026-11-16): 977 SNV and 54 indel residues; the loader replaces only the sheet it is given.
  • CIViC (evidence items and oncogenicity assertions) (2026-11-16): Nightly release of 2026-11-16: 1,723 rows on 99 main isoforms (single-residue missense changes and residue-level profiles); feeds the somatic reading's OS1 and OM4.
  • Browse snapshot (precomputed browse tables) (2026-11-16): static/precomputed_browse/manifest.json.
November 2026

Explore, Statistics, somatic reading, Your data and one page structure

What changed since April, by page.

Summary › Explore

  • One Tracks drawer in every view replaces the Advanced panel: switch tracks on and off under the figure's own section headings, split or merge rows, reorder them, and keep a separate choice for each view. Presets (disorder-, variants- and motifs-focused) and Copy link, which carries the view, tracks, order and regions in the address.
  • Each view has its own settings row. Structure: PDB | AlphaFold, an entry picker, Mark entry, Water & ligands and chips that mark every ordered, disordered or complex-bound region at once. Variant effect: Functional map | Compare predictors and Marks on map. Orthologs: taxonomic level and species, with silhouettes.
  • What stands out above the figure, and significantly mutated regions as one merged strip under the variants. The Overview is the protein at a glance: subcellular location, affected tissues, variants by source and context, with All · Somatic · Germline pathogenic.
  • Custom regions take their own colour and can be named on the track. Export saves the figure as PNG or SVG, with size presets for journal single and double columns, and can name the highlighted regions.
  • The same Explore figure is now embedded in Variant interpretation (Protein context) and in AI Chat (Tracks view).

Summary › Statistics and Tables

  • Statistics is a dashboard of nine cards in which every count is set against a baseline (the protein's own residues in the same structural context, or the proteome), with Origin and Scope filters and predictor diagnostics against germline pathogenic variants and the MAVE assays.
  • Tables is a tab of its own and builds each table when you open it.

Variant interpretation

  • One row (protein, variant, Interpret) with the older steps under Advanced. It also takes genomic input: an rsID, 7-140753336-A-T, a ClinVar VCV id or an HGVS name, in GRCh38 or GRCh37 (since 2 November).
  • A report with Overview (an evidence ledger whose rows are never combined), Evidence, Classification (automated ACMG/AMP criteria with gnomAD 4.1 frequency, since 8 October), Protein context and Sequence context. A mismatched reference residue stops the run instead of being interpreted.
  • A somatic reading beside the germline one, following the ClinGen / CGC / VICC oncogenicity standard, with cancerhotspots.org (OS3, OM3, OP3) and CIViC (OS1, OM4) and a cancer-evidence block. The two readings are never merged.

AI Chat

  • Example workflows and answers with one primary view and a More views list; germline and somatic readings side by side; phase separation shown as protein-level context; citation chips and a Sources list; each answer is checked against the data it used.

Analysis

  • Overview, Compare a set and Your data. Compare a set opens on a table, each row set against a baseline, with a map of expected against observed; Simple is three choices, Advanced the full set of controls.
  • Your data reads proteins, regions or variants, including a VCF or MAF in GRCh37 or GRCh38, and keeps them in your browser tab, not on the server.

Navigation and wording

  • Home starts from six tasks with an example each, and one search box that sends a gene to its Summary, a variant to Variant interpretation and a GO term or disease to Compare a set. Answers end in Next steps links that carry the protein, variant or set to the next page.
  • One page structure on every page (menu, page header, one content width), a Data & API menu (Downloads, API) and a Help menu (Help, About & cite).
  • Wherever a variant's origin is named it says Germline or Somatic, with the source beside it where one source is meant. Spelling is British.

Data

  • New: CIViC evidence and oncogenicity assertions and cancerhotspots.org v2 (both 16 November), the MoMAP motif release of 9 October (loaded 3 November) and the GRCh37 / GRCh38 genome map behind genomic input (3 November).
  • Reloaded 16 November: PDB spans (now 1-based, from SIFTS), coiled coils (from our DeepCoil run, 1-based) and the precomputed Browse tables. The list is under Rebuilt in this release above; every source, version and date is under Data sources.
April 2026

Full UI rebrand — DisCanVis 2.0

  • End-to-end visual refresh: Bootstrap-based design system, brand colours, dark-mode–aware Summary (Plotly, cards, sequence toolbar), loaders, and compact navigation.
  • Home page: DisCanVis 2.0 positioning, link to legacy v1.discanvis.elte.hu, search without the old narrow-search gate, refreshed examples (transcript ID, long gene name), live main-isoform database counters, disorder-dataset explorer (MFIB, DIBS, PhasePro, ELM, PEM, …), and a grouped Databases & methods strip (colour keys aligned with Summary / Browse).
  • Footer & About: shorter footer, unified citation formatting, release timeline with calendar months (this page).
  • Continued polish on browse flows, Batch Analysis entry points, and Statistics where applicable.
February 2026

Structure viewer & Summary tracks

  • PDBe Mol* integration and viewer assets revised, with cleaner loading and layout next to the main Summary plot.
  • Refinement of sequence tracks, mutation layers, and disorder / annotation overlays aligned with the updated viewer stack.
December 2025 – January 2026

Search results & protein triage

  • Unified search results (home, navbar, /search): DataTables table, filters, pagination, Copy / CSV / Excel / Print.
  • Rows show ordered vs disordered composition, the somatic vs germline mix, and the per-database variant spectrum (missense, frameshift, indel) for TCGA, COSMIC, cBioPortal, etc.
  • Shortcuts to Summary, PPI, and REST JSON.
November 2025

Pathogenicity predictors on Summary

  • Integrated dbNSFP-style ensemble pathogenicity tracks and AlphaMissense alongside existing mutation and disorder layers in the Summary visualisation.
  • Dropdowns and colouring options updated so predictors stay in sync with the sequence view and feature table.
2023 · Data & browse (v1.2 era)

Earlier milestone

  • Expanded somatic mutation coverage (COSMIC, TCGA / legacy TCGA, cBioPortal as deployed).
  • PPI from IntAct, BioGRID, HIPPIE.
  • Extended browse tables (e.g. OMIM diseases, phase separation annotations).

Funding & partners

DisCanVis is supported by Eötvös Loránd University, with additional support from the following partners.

IDP2Biomed
IDP2Biomed

Twinning of research institutions (EU programme).

DisCanVis is part of the ELIXIR intrinsically disordered proteins community and aligns with the broader ELIXIR infrastructure.