is an Aggregator.
Sugi Atlas is a deterministic biomedical reference atlas with one canonical page for each human gene, drug and disease (about 52,600 pages when checked on 2026-10-04, 29,337 genes, 4,683 drugs and 18,607 diseases). Pages are assembled by a fixed plan of graph queries over BioBTree, which integrates 87 primary databases, and are rendered by template, with no language model in the loop, so every figure traces to its source. Each page is published as static HTML with a Markdown version and a schema.org JSON-LD record, for both human readers and AI agents. Gene, drug and disease pages are seeded from HGNC, Mondo and ChEMBL identifiers and are rebuilt at least monthly against new BioBTree releases.
No external IDs available
| ID | Name | URL | Category | Format | Description |
|---|---|---|---|---|---|
| sugi-atlas.portal | Sugi Atlas Web Site | atlas | Graphical Interface | http | Sugi Atlas web site with search and o... |
| sugi-atlas.manifest | Sugi Atlas Manifest | manifest.json (12.2 MB) | Product | json | JSON manifest of the atlas corpus, ma... |
| sugi-atlas.code | Sugi Atlas Pipeline | sugi-atlas | Process | python | Python pipeline that queries BioBTree... |
| sugi-atlas.preprint | Sugi Atlas Preprint (HTML) | preprint | Documentation | http | HTML version of the Sugi Atlas prepri... |
| sugi-atlas.sources | Sugi Atlas Data Sources | sources | Documentation | http | List of the 87 primary databases inte... |
Sugi Atlas is a biomedical reference atlas: one canonical, structured page for every human gene, drug and disease. It is part of the sugi.bio project and is built on BioBTree, which integrates the 87 primary databases listed on the Sources page.
Every page is built deterministically. A fixed plan of graph queries traverses BioBTree, and the returned records are rendered into tables and prose by template. No part of a page is written by a language model. The same inputs give the same page on every build. Where a section has no data, the page says so.
Seeds are pre-resolved HGNC (genes), Mondo (diseases) and ChEMBL (drugs) identifiers. The drug set covers approved (phase 4) and late-stage clinical (phase 3) compounds. Non-human and veterinary disease terms are dropped at seed time.
Genes, drugs and diseases are linked by a curated cross-entity mesh, so a relationship recorded once (a drug’s target, a disease’s gene) is navigable from every page it touches.
https://sugi.bio/atlas/{gene,drug,disease}/<slug>/, for example TP53..../index.md, with a metadata header.No bulk download of the corpus is published. The pipeline releases on GitHub version the code only.
The pipeline code is MIT. The generated atlas content is published under CC BY-SA 4.0, according to the repository README, and each underlying source keeps its own license and attribution terms.
Gür T. Sugi Atlas: A Comprehensive, Deterministic Biomedical Catalog from a Knowledge Graph for Humans and AI Agents. Zenodo, 2026. doi:10.5281/zenodo.20569200
Created: October 04, 2026 | Last modified: October 04, 2026