POTS deep phenotyping

The literature layer for an OMOP computable phenotype of postural orthostatic tachycardia syndrome: a harvested PubMed corpus, a topic vocabulary, and an ontology linking tests, treatments and conditions to the possible causes of POTS, each link backed by a quoted passage.

Draft for review. Each link was found by AI readers: Claude and at least one System One model (Jev, winnow or laya) had to find it in the same paper. No clinician has reviewed the links yet.
2,668Articles
431Ontology edges (taxonomy included)
72Facets
329Literature links (all three readers found it, Claude confirmed it)
3098Papers backing them (one quote each)
42Links for review

Explore

Evidence map
The ontology as a DAG rooted at POTS, a map, an outline and every link with the quoted passage behind it.
Interactive
Reader votes
Where Claude and the System One models (Jev, winnow, laya) agreed and disagreed on each link, six cases read closely with the exact questions and answers, and every vote.
Interactive

Reports

From seed papers to an ontology of POTS
Everything so far: the seed papers, the vocabulary, the four readers, the vote, and the ontology as it stands.
PDF
Rebuilding the ontology from seven seed papers
How the vocabulary, the evidence links, the corpus and the concept mappings were rebuilt evidence first, and what came out.
PDF
Can the AI reading step be reproduced?
The seven seed papers read three times each by a scripted claude -p reader: how far the readings agree, and with the ontology.
PDF

Documentation

Competency questions
The questions the ontology answers, one per relationship type.
Markdown
The relation graph
Predicates, why the graph is shaped this way, and how to query it.
Markdown
Harvest report
Generated validation report for the committed run.
Markdown
Data dictionary
Every table and column in the database.
Markdown
README
Project overview, design decisions and quick start.
Markdown

Data