The GENCODE protein accession, as in a Summary URL (RAF1-201). /rest/ and /download/ also take a gene symbol, a UniProt accession or an Ensembl transcript with search_type; a gene with several isoforms answers with its main isoform.
Positions
1-based and inclusive, on the protein sequence. The one exception is the pdb layer of /rest/: its spans are 0-based (Summary shifts them by one).
Release
Every JSON answer of /rest/ has a meta block: api_version, generated and release (data release 2026.11, genome, link to the source list). Cite the release with the data you use.
Errors
/rest/ answers a wrong protein, layer or range with HTTP 200: a text message, or an error key in the JSON. Look for it. The other endpoints use HTTP 400, 404 or 503 with an error key.
Stability
/rest/, /download/ and /api/v1/ are the documented interfaces. Protein reports are what the pages draw and can change with them: read the field you need by name.
Use
No key, no registration. Ask for one protein, region or residue at a time; for the whole proteome use Downloads. Everything here is read-only.
travel_exploreFind a protein
Turn a name you have into the protein identifier the rest of the API takes.
GET/find/suggest
Search suggestions
What the search box offers for what you typed: where it would go (intent) and grouped suggestions.
Parameters
Name
In
Description
q
query · required
A gene, accession, variant, GO term, disease or a sentence.
/find?q=… is the same lookup as a redirect (HTTP 302) to the page that answers it; a script should read this endpoint and take the accession from intent.url. An empty q answers {"intent": null, "groups": []}.
One protein: every layer, one layer, one region or one residue. Positions are 1-based.
GET/rest/{id}.{format}
A whole protein
Every layer for one protein: JSON (a meta block and one key per layer), tab-separated text, or FASTA.
Parameters
Name
In
Description
id
path · required
The protein, as named by search_type.
format
path · required
json, txt (one #layer section per layer) or fasta (>accession|UniProt|transcript|gene).
search_type
query
What id is: empty = GENCODE protein accession (RAF1-201); gene_name (RAF1), uniprot (P04049) or transcript (ENST00000251849.9). A gene with several isoforms answers with its main isoform.
Large: the JSON of RAF1-201 is about 1.6 MB. Read one layer, region or residue when you can. A layer with no rows is null. pdb spans are 0-based here (Summary shifts them by one).
One layer of one protein as tab-separated text; the first column is a row number.
Parameters
Name
In
Description
id
path · required
The protein, as named by search_type.
layer
path · required
A layer name from the table at the foot of the page.
search_type
query
What id is: empty = GENCODE protein accession (RAF1-201); gene_name (RAF1), uniprot (P04049) or transcript (ENST00000251849.9). A gene with several isoforms answers with its main isoform.
The answer is text wrapped in <pre>. A wrong layer or protein is still HTTP 200, with the message as the text (the list of valid layers, or Not valid Protein ID).
The layers inside residues start to end (1-based, inclusive).
Parameters
Name
In
Description
id
path · required
The protein, as named by search_type.
start
path · required
First residue.
end
path · required
Last residue; not beyond the sequence length.
format
path · required
json, or txt for tab-separated text.
annotations
query
A JSON list of the layers to keep, e.g. ["clinvar","tcgam"]; empty = every layer. Only the layers marked yes under Layer names exist for a region or residue. The disorder layers answer under one key, Disordered.
search_type
query
What id is: empty = GENCODE protein accession (RAF1-201); gene_name (RAF1), uniprot (P04049) or transcript (ENST00000251849.9). A gene with several isoforms answers with its main isoform.
A JSON list of the layers to keep, e.g. ["clinvar","tcgam"]; empty = every layer. Only the layers marked yes under Layer names exist for a region or residue. The disorder layers answer under one key, Disordered.
search_type
query
What id is: empty = GENCODE protein accession (RAF1-201); gene_name (RAF1), uniprot (P04049) or transcript (ENST00000251849.9). A gene with several isoforms answers with its main isoform.
Errors come the same way as for a region: HTTP 200 with an error key. annotations=["roisig"] is not usable here: that name is answered with the disorder layers.
The /rest/ answers with a download header, for a browser or wget.
GET/download/{id}.{format}
A whole protein, as a file
The same answer as /rest/{id}.{format}, with a Content-Disposition header that saves it.
Parameters
Name
In
Description
id
path · required
The protein, as named by search_type.
format
path · required
json, txt or fasta.
search_type
query
What id is: empty = GENCODE protein accession (RAF1-201); gene_name (RAF1), uniprot (P04049) or transcript (ENST00000251849.9). A gene with several isoforms answers with its main isoform.
One layer as a tab-separated file with a header row (no row-number column).
Parameters
Name
In
Description
id
path · required
The protein, as named by search_type.
layer
path · required
A layer name from the table at the foot of the page.
search_type
query
What id is: empty = GENCODE protein accession (RAF1-201); gene_name (RAF1), uniprot (P04049) or transcript (ENST00000251849.9). A gene with several isoforms answers with its main isoform.
The same answer as /rest/{id}/{position}.{format}, as a saved file.
Parameters
Name
In
Description
id
path · required
The protein, as named by search_type.
position
path · required
The residue.
format
path · required
json or txt.
annotations
query
A JSON list of the layers to keep, e.g. ["clinvar","tcgam"]; empty = every layer. Only the layers marked yes under Layer names exist for a region or residue. The disorder layers answer under one key, Disordered.
search_type
query
What id is: empty = GENCODE protein accession (RAF1-201); gene_name (RAF1), uniprot (P04049) or transcript (ENST00000251849.9). A gene with several isoforms answers with its main isoform.
Nothing is stored. The answer is evidence, not a clinical classification. If the reference residue is not the residue in the sequence, ref_mismatch is true and acmg is null. A single variant takes 2 to 3 s. Unknown protein or no position is HTTP 400 with ok: false.
Large and slow (RAF1-201: 3.6 MB, about 4 s). Without accession: HTTP 400 missing_accession; unknown accession: 404. The structural shortlist reports co-occurrence in a PDB entry, not a computed interface.
<!DOCTYPE html> … the Explore figure with its View / Settings bar, for an iframe … (HTML, not JSON)
Also compare=1 with predictors=a,b (Variant effect with the predictor lanes compared). /summary/embed/overview/{accession}/, …/structure/… and …/table/… are the Overview, the structure view and the feature table in the same way. Unknown accession: 404 Protein not found.
The same filters as the table (region_sets, sources, chromosome, go, disease, patho_*, …; the catalog lists them). An unknown cohort key is not an error (HTTP 200): read cohorts in the answer for what was used.
An unknown columns value is HTTP 400 naming the valid ones (unknown column(s): …); an unknown format is 400 too. Also mode, q, cf_<column> filters and the filters of the summary.
A saved set is a snapshot file (human: 2.7 MB). The others are computed and refuse a set that is too large (HTTP 400); a missing snapshot is 503. Unknown set: 400.
Kept for scripts that already call it; new code should use /api/v1/analysis/scan/. GET answers a short description of the endpoint. Empty roi: HTTP 400 {"success": false, "errors": […]}.
The names /rest/{id}/{layer} and /download/{id}/{layer} take, read from the code. A region or a residue answers the layers marked yes; the others exist only for a whole protein. A layer with no rows is empty or null.
Predictor scores per substitution (dbNSFP-style columns, AlphaMissense).
no
pdb
PDB structure spans, ordered and disordered stretches. Coordinates are 0-based here.
yes
pfam
Pfam domains with alignment and envelope spans.
yes
combined
Combined disorder score per residue (disordered at 2.0 or more).
yes
anchor
ANCHOR2 binding propensity.
yes
iupred
IUPred3 disorder score.
yes
aiupred-binding
AIUPred-Binding score.
no
mobidb
MobiDB experimental disorder.
yes
alphafold
AlphaFold pLDDT confidence.
yes
tcgam
Somatic missense variants from TCGA, with cancer types.
yes
tcgaf
Somatic frameshift variants from TCGA.
yes
tcgai
Somatic in-frame indels from TCGA.
yes
cosmicm
COSMIC missense (COSMIC is not loaded: no rows).
yes
cosmicf
COSMIC frameshift (not loaded: no rows).
yes
cosmici
COSMIC indel (not loaded: no rows).
yes
cbioportalm
Somatic missense variants from cBioPortal.
no
cbioportalf
Somatic frameshift variants from cBioPortal.
no
cbioportali
Somatic in-frame indels from cBioPortal.
no
roisig
Significantly mutated regions (iSiMPRe).
yes
elm
ELM motif instances.
yes
pem
PEM predicted motifs.
no
momap
MoMAP motifs.
no
scansite
ScanSite kinase and binding motifs.
yes
elmswitches
ELM switches (a binding site and the modification that gates it).
yes
ptm
Post-translational modification sites.
yes
roi
UniProt regions of interest.
yes
binding
UniProt binding sites.
yes
dibs
DIBS disordered binding regions.
yes
mfib
MFIB fuzzy complexes.
yes
binding_domain
Binding domains.
yes
phasepro
PhaSePro phase-separation regions.
yes
conservation
Conservation scores per taxonomic level.
yes
ppi
Interaction partners (HIPPIE, IntAct, BioGRID).
no
In a region or residue answer the disorder layers (combined, anchor, iupred, mobidb, alphafold) share one key, Disordered. pathogenicity holds the predictor scores per substitution, AlphaMissense among them; it is not in a region or residue answer.
codeFrom Python
Set BASE to this site or another deployment.
The variants in a region
import requests
BASE = "https://discanvis.elte.hu"
r = requests.get(
f"{BASE}/rest/RAF1-201/250-265.json",
params={"annotations": '["clinvar", "tcgam"]'},
timeout=60,
)
data = r.json()
if "error" in data: # /rest/ answers errors with HTTP 200
raise SystemExit(data["error"])
print(data["meta"]["release"]["data_release"], data["tcgam"]["mutations"])
A set of proteins as a table
import io
import pandas as pd
import requests
BASE = "https://discanvis.elte.hu"
r = requests.get(
f"{BASE}/api/v1/analysis/table/",
params={"cohorts": "cancer_drivers", "unit": "protein", "format": "csv",
"columns": "gene,protein,missense,disordered_pct", "limit": 200},
timeout=120,
)
r.raise_for_status() # these endpoints use 4xx with an "error" key
df = pd.read_csv(io.StringIO(r.text))