Skip the thousand PDFs. Open source, research only

Run your first PRNP analysis.

Point it at PubMed and get a continuously queryable dataset: every reported variant (E200K, D178N, P102L…) with its reported association, evidence provenance, and CSV + Markdown export. Not a classifier: cross-check every call against ClinVar.

Presets

The pipeline, end to end

every step is deterministic and inspectable. No black box.

1
Query PubMed
esearch.fcgi via public E-utilities (no API key)
2
Retrieve papers
efetch.fcgi pulls title, abstract and metadata per record
3
Detect variants
Regex scanners: E200K · Glu200Lys · OPRI/OPRD
4
Normalize
3-letter → 1-letter codes; codon window 40–243 validated
5
Match KB
~45 curated PRNP entries (ClinVar / UniProt / literature)
6
Annotate evidence
Reported association + evidence tier + provenance
7
Export
CSV + Markdown via /api/export, or grab the Python script

Live example output

What you get when you click Analyze

loading corpus…

Input: the literature

A wall of abstracts lands here the moment the corpus is seeded.

Reading 100 abstracts by hand ≈ 2–3 h. The pipeline below does it in seconds, reproducibly.

Output: the structured variant/evidence dataset

Variant	Paper	Syndrome reported	Evidence	PMID

tiers describe what the literature reports, not clinical validationcross-check ClinVar ↗

Export structure: variants CSV

/api/export?type=variants

…

Papers CSV too (pmid, journal, title, doi, variants detected, PubMed URL); download buttons live in the dataset section below.

Frequency report: literature mentions per variant

Report tab · .md

| Variant | Mutation Type | Number of Mentions |

Counts appear as soon as the corpus is seeded.

Every textual mention across titles + abstracts is counted (same table the bundled Python script prints).

Evidence tiersDisease association reported in literatureFunctional / experimental evidencePopulation-level associationMachine-detected, not yet annotated

Zero-setup playground

Paste any text. Watch it become structured data

The exact same extraction engine as the corpus pipeline (regex scanners, codon-window guard, knowledge-base match, evidence tiers), running right here. No PubMed query, no database write.

abstract, case report, review paragraph…

deterministic · inspectable · nothing is stored

Beyond PRNP: one connected field

scrapie, kuru, BSE and CJD became one disease family because researchers connected observations across disciplines. Live PubMed corpus sizes for the fields around this tool.

Each card runs its term through the full analyzer: the same pipeline, just pointed at a neighbouring field. Counts are cached for an hour so E-utilities stay rate-limit friendly.