Skip to content

Route A — Manual via the DOME Wizard

You, the paper, and 21 fields. The slowest route, but the one that teaches you the most.

Overview

Questions:

  • How do I create a DOME entry by hand?
  • Do I need an account, and how long does that take?
  • What do I do about fields the paper simply does not answer?

Learning Objectives

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

  1. Sign in to the DOME Wizard, with LS Login or a new account
  2. Create a project from the DOME knowledge model
  3. Fill in the Data and Model sections for a real paper
  4. Record what the paper does and does not report, honestly

Time: 35 minutes (hands-on exercise)

Prerequisite: Chapter 1 for the field definitions · an LS Login or DOME Wizard account

Which paper should I annotate?

Your own, if you have one with a supervised ML method — you will get the most out of it. Otherwise pick any supervised ML paper in biology you know reasonably well. A paper you have reviewed works especially well.

Need one? Browse finished entries in the DOME Registry search, open any entry, and follow its DOI through to the paper.

Annotating in the DOME Wizard

The DOME Wizard is the DOME questionnaire running on ELIXIR’s Data Stewardship Wizard: a guided, question-by-question interview with per-question guidance, and comments and TODOs you can leave on individual answers.

Steps

  1. Go to https://dome.dsw.elixir-europe.org/wizard/

  2. Sign in. If you already have LS Login, use that button — it is the smoothest way in. Otherwise create a DOME Wizard account: the confirmation email usually arrives within a minute or two, and you can log in as soon as you have clicked the link in it.

  3. Projects → Create → give it any name (MAQC 26 Test is fine) → pick the DOME knowledge model → Create.

  4. Pick your paper — your own, or one you found in the Registry.

  5. Fill in the Data section first, then Model, then as much of the rest as the time allows. Keep Chapter 1 open in another tab for the field definitions.

After the session

For a real annotation, the Wizard is what sends your finished entry on to the DOME Registry. Today, filling it in is the exercise.

Working through the fields

Work in this order — it follows how a paper is usually written, and the Optimisation fields are the fiddliest, so they are best left until last:

Pillar Where the answers usually live
Data Methods, “Data availability” statement, supplementary tables
Model Methods, “Code availability” statement, the repository itself
Evaluation Results, figures and their captions, supplementary benchmarks
Optimisation Methods, supplementary methods, the code repository’s README or config files

Record gaps as gaps

When the paper does not state something, leave the field empty and note it. Do not fill it from the GitHub repository, from a previous paper by the same group, or from what you assume they must have done. DOME measures what the publication discloses — inferring the answer defeats the entire exercise.

(The agent skill handles this by tagging any repo-sourced field as external:<url> and flagging it as not disclosed in the paper itself. Do the same thing by hand: note it separately.)

If you get stuck on a field

Go back to Chapter 1 and read the “why it matters” line for that field. It usually makes clear what evidence would satisfy it — and, just as often, makes clear that the paper genuinely does not provide it.

When to choose this route

Good fit: your own paper; you want to learn the recommendations properly; small numbers; no LLM in the loop for policy or preference reasons; maximum control and confidence in every field.

Poor fit: more than a handful of papers. At 30–90 minutes each, manual annotation does not scale — which is the entire reason Route B and Route C exist.

Next: Route B — DOME Copilot