pampalini1/olivers-mtor-atlas

Oliver's mTOR Atlas

The mTOR pathway, mapped by what the evidence can actually carry. This dataset is the curated corpus behind mtor-atlas.org: 411 hand-selected studies on mTOR (mechanistic target of rapamycin) signalling, each labelled by the kind of study behind it, and a list of 149 pathway entities (genes and proteins, complexes, drugs, interventions, biological processes, diseases, outcomes, organelles, nutrients and conditions) that the studies refer to.

• —Homepage: https://mtor-atlas.org
• —Curator: Oliver Barton (ORCID 0009-0008-2025-2148)
• —License: CC BY 4.0
• —Dataset DOI (Zenodo, all versions): 10.5281/zenodo.22059963
• —Companion paper (Figshare): "Ten percent human: an evidence-graded audit of the mTOR literature and the pathway's translational gap", 10.6084/m9.figshare.33772297
• —Also registered with: bio.tools, FAIRsharing, Wikidata Q141256074

Files

• —studies.csv / studies.json: one row per study. Atlas ID (sid), title, authors, year, journal, study type (tier, pyramid), category and model system, DOI/PMID/PMCID, a one-line curated finding, the PubMed abstract and, where extracted, AI-assisted fields (intervention, target, species, effect, dose, sample size, effect size, limitations). atlas_url links to the record page.
• —entities.csv / entities.json: one row per entity. Name, type, a technical and a plain-language description, synonyms, the number of linked studies and, for entities with their own page, atlas_url.

How the study-type labels work

The website shows five codes: S (synthesis of human data), H (human study), A (animal model), M (molecular or in vitro work) and R (narrative review). They say what kind of evidence a study is. They are not a quality ranking, and a careful animal study is not "worse" than a weak human one.

The tier column keeps the original stored letters, so read it through this mapping:

A few rows carry Preprint or Registered trial instead.

Intended uses

• —Evaluating biomedical QA or summarisation models on a small corpus where every row traces back to a DOI or PMID.
• —Testing study-type classifiers against human curation.
• —Knowledge-graph work on the mTOR pathway, with the entities file as a node list.

Limitations

This is a curated selection, not a systematic review of the whole mTOR field, so a paper's absence says nothing against it. The label describes study design, not effect size or statistical power. The corpus grows as new studies are added, and this Hub copy may lag the live site. For a fixed, citable snapshot use the Zenodo DOI. Current exports are always at https://mtor-atlas.org/data/.

Citation

@misc{olivers_mtor_atlas,
  author    = {Barton, Oliver},
  title     = {Oliver's mTOR Atlas},
  year      = {2026},
  publisher = {Zenodo},
  doi       = {10.5281/zenodo.22059963},
  url       = {https://doi.org/10.5281/zenodo.22059963}
}