PhaSePred API Documentation

PhaSePred is a comprehensive resource for predicting liquid-liquid phase separation (LLPS) related proteins. This API provides programmatic read-only access to all prediction results.

• Base URL: http://predict.phasep.pro/api/v1/ (via public IP the same API is reachable at http://47.88.20.47/phasepred/api/v1/)
• Data version: 2022-02-11 — 116,806 reviewed proteins across 20 species
• Authentication: none (public read-only)
• Rate limit: 10 requests/second per IP with burst 20; exceeding it returns HTTP 503
• All responses are JSON. Success: {"data": ...}. Error: {"error": {"code": ..., "message": ...}}

Endpoints

Endpoint	Method	Description
/api/v1/	GET	Service index (version, endpoint list)
/api/v1/meta/	GET	Dataset statistics, organism list, tool descriptions
/api/v1/protein/<entry>/	GET	Full prediction results for one UniProt entry
/api/v1/search/	GET	Keyword search
/api/v1/proteins/batch/	GET, POST	Batch retrieval (up to 100 entries)
/api/v1/downloads/	GET	Per-species bulk download links

GET /api/v1/protein/<entry>/

Full prediction results for a single protein. <entry> is a UniProt accession (case-insensitive, e.g. O14958).

Response fields inside data:

Field	Type	Description
entry	string	UniProt accession
entry_name	string	UniProt entry name (e.g. CASQ2_HUMAN)
status	string	UniProt review status (reviewed)
gene_names	string	Gene names
organism	string	Organism
sequence	string	Protein sequence
domain	object	Pfam domains: PfamID[], domain[], start[], end[]
phasepred	object	PhaSePred meta-predictor scores: SaPS-8fea, PdPS-8fea, SaPS-10fea, PdPS-10fea and *_rnk ranks
rank	object	rank[] and method[] (12 methods)
tools	object	Per-tool results, see below

tools sub-fields: catgranule, plaac, pscore, espritz, hydropathy, deepcoil, seg, charge, phos, deepphase. Each contains residue-level score arrays (per-residue values as JSON number arrays), summary scores (single/NLLR/FCR), region boundaries (start[], end[]) and organism rank (rnk).

Example:

curl -s http://predict.phasep.pro/api/v1/protein/O14958/

{
  "data": {
    "entry": "O14958",
    "entry_name": "CASQ2_HUMAN",
    "organism": "Homo sapiens (Human)",
    "phasepred": {"SaPS-8fea": 0.43, "PdPS-8fea": 0.60, "...": "..."},
    "tools": {
      "catgranule": {"residue": [0.0, 0.0, 0.27, "..."], "single": 0.12, "start": [1], "end": [45], "rnk": 0.31},
      "...": "..."
    },
    "domain": {"PfamID": ["PF01216"], "domain": ["Calsequestrin"], "start": [2], "end": [381]},
    "sequence": "MALLHSAR..."
  }
}

Errors: 404 not_found if the entry does not exist.

GET /api/v1/search/

Parameter	Required	Default	Description
q	yes	—	Keyword, at least 2 characters
condition	no	Protein_Name	Protein_Name (matches gene/entry names) or Uniprot_ID (matches UniProt accessions)
limit	no	20	Page size, 1–100
offset	no	0	Pagination offset

Response data: q, condition, limit, offset, has_more (boolean), and results[] with status, entry, entry_name, organism, gene_names.

curl -s "http://predict.phasep.pro/api/v1/search/?q=CASQ2&condition=Protein_Name&limit=5"

GET or POST /api/v1/proteins/batch/

Retrieve up to 100 proteins in one request.

• GET: ?entries=O14958,P12345
• POST: JSON body {"entries": ["O14958", "P12345"]} with Content-Type: application/json

Entries are case-insensitive and de-duplicated. Response data: count, results[] (same shape as the single-protein endpoint) and not_found[].

curl -s "http://predict.phasep.pro/api/v1/proteins/batch/?entries=O14958,P12345"
curl -s -X POST -H "Content-Type: application/json" \
     -d '{"entries": ["O14958", "P12345"]}' \
     http://predict.phasep.pro/api/v1/proteins/batch/

Errors: 400 bad_request if entries are missing, malformed, or exceed 100.

GET /api/v1/meta/

Dataset statistics: data_version, total_proteins, organisms[] (name + protein count, 20 species) and tools (description of every tool field).

GET /api/v1/downloads/

Lists per-species bulk download packages. Each item: file, organism, url (relative to the site root, e.g. /static/phasepred/database/human_reviewed.zip), size_bytes. 18 packages are available.

GET /api/v1/

Service index: name, API version, data version and the endpoint list.

Error codes

HTTP	error.code	Meaning
400	bad_request	Missing/invalid parameters or malformed JSON body
404	not_found	Entry does not exist
405	method_not_allowed	Wrong HTTP method
503	—	Rate limit exceeded (nginx)

Python example

import requests

BASE = "http://predict.phasep.pro/api/v1"

r = requests.get(f"{BASE}/search/", params={"q": "CASQ2", "limit": 5})
entries = [item["entry"] for item in r.json()["data"]["results"]]

r = requests.post(f"{BASE}/proteins/batch/", json={"entries": entries})
for protein in r.json()["data"]["results"]:
    print(protein["entry"], protein["phasepred"]["SaPS-8fea"])

Citation

Chen Z, Hou C, et al. Screening membraneless organelle participants with machine-learning models that integrate multimodal features. PNAS 2022;119:e2115369119.

Terms

Free for non-commercial use for academic, government and non-profit institutions.