Skip to content

2. The DOME Registry

A public home for DOME transparency reports — so they are findable, shareable, citable and comparable.

Overview

Questions:

  • A DOME report is useful — but where does it live, and who can find it?
  • How does a registry entry help an author, a reviewer, an editor and a meta-researcher differently?
  • What is a DOME score, and what can it legitimately be used for?

Learning Objectives

By the end of this chapter, you will be able to:

  1. Explain why DOME reports need a registry rather than living in supplementary PDFs
  2. Describe how a registry entry slots into a journal’s peer-review workflow
  3. Interpret a DOME score, and state its limits

Time: 10 minutes

Prerequisite: Chapter 1 — The DOME Recommendations

The problem a registry solves

A DOME report buried in Supplementary Table 7 of a PDF is, for practical purposes, invisible. It cannot be searched, compared, updated, linked to from a review, counted across a field, or reused by anyone building tooling on top of it. The reporting effort is spent and then lost.

The DOME Registry is the answer: a public, structured, versioned database of DOME transparency reports, described in GigaScience 1. Entries are stored as structured JSON against a controlled schema, given persistent identifiers, and linked to their publication and to the curator who created them.

The shift

From “the DOME checklist is a thing you fill in once” to “the DOME report is a research object with an identifier, a version history and an author.”

Who it serves, and how

Complete a DOME report during manuscript preparation, get a stable link, and supply that link at submission. It is a concrete, low-effort way to demonstrate methodological transparency — and filling it in usually improves the methods section itself.

Open one link and see, field by field, what the paper does and does not report. As the GigaScience paper puts it, reviewers gain a standardised and efficient evaluation tool — instead of reconstructing the methods from prose, you are checking a structured record against the manuscript.

Integrate the report into the submission workflow so that ML methodology is assessed consistently across submissions. The benefit that matters editorially is improved submission quality: authors who know they will be asked tend to report better.

Structured entries can be counted, queried through the REST API, and analysed across a field — which is what makes questions like “has data-split reporting improved since 2021?” answerable at all.

What the Registry provides

Feature What it does
LS Login authentication Sign in with LS Login, the European life-science AAI; curation is attributed to a real, persistent identity
DOME score Automated compliance score, 0–21, one point per completed field
Versioned entries Entries can be revised; earlier versions remain interpretable against their schema version
Moderation workflow Drafts can be submitted for review before appearing publicly
REST API Programmatic read and write access — what tooling like DOME Copilot and the DOME Agent Skill build on
APICURON integration Biocuration contributions are credited, so curation effort is visible and countable
Data Stewardship Wizard The DOME questionnaire is also available through ELIXIR’s DSW instance
Zenodo archival The full dataset is versioned and backed up to Zenodo automatically

Reading a DOME score honestly

The score is 0–21: one point per DOME field that has been completed.

What the score is not

The DOME score measures how much was reported, not how good the science is. A well-reported weak method scores higher than a brilliantly executed method whose authors said nothing about their data splits. Use it as a transparency indicator and a prompt for specific questions — never as a proxy for methodological quality, and never as a ranking of researchers.

Used well, the score is most valuable comparatively: across a journal’s submissions, across a subfield, or across versions of your own work as you improve its reporting.

Sources and further reading

Next: Chapter 3 — DOME Copilot

  1. Omar Abdelghani Attafi, Davide Cirillo, Alexander Miguel Monzon, Maria Chatzou Dunford, Gavin Farrell, Jennifer Harrow, Fotis E. Psomopoulos, and Silvio C. E. Tosatto. DOME registry: implementing community-wide recommendations for reporting supervised machine learning in biology. GigaScience, 13:giae094, 2024. doi:10.1093/gigascience/giae094. ↩↩