Publishing Reusable AI/ML Models in the Life Sciences
MAQC 2026 Technical Tutorial P06
Wednesday 23 September 2026, 14:00–15:30 · Brescia, Italy
Led by Gavin Farrell, University of Padua (BioComputingUP)
Open the tutorial slides Session assets
Overview
Questions this tutorial answers:
- What does it actually take to report an AI/ML method transparently enough that someone else can trust and reuse it?
- What are the DOME recommendations, and what do their 21 fields ask for?
- How do the DOME Registry, DOME Copilot and agentic AI skills each help me produce a DOME transparency report?
- How does OSAI extend this from reporting a method to sharing and sustaining a model?
- Which route should I use for my own paper — and what does each one cost me in time, money and control?
Learning Objectives
By the end of this session, you will be able to:
- Explain the four DOME pillars — Data, Optimisation, Model, Evaluation — and identify which of the 21 fields a given paper fails to report
- Describe why a public registry of DOME reports matters for authors, reviewers, editors and meta-researchers
- Produce a DOME transparency report for a paper by at least one of three routes: manual, DOME Copilot, or an agentic AI skill
- Position the nine OSAI recommendations alongside DOME and identify concrete ecosystem components that close a reporting gap
- Judge the trade-offs between the three routes — accuracy, cost, accessibility and required human review
Who this is for: Researchers publishing AI/ML methods, data curators, journal editors and peer reviewers, and anyone assessing the reproducibility of computational methods in the life sciences
Prerequisites: Familiarity with supervised machine learning terminology. No coding is required for Routes A and B; Route C needs Python 3.10+ and a terminal
Time: 90 minutes
Why this session exists
AI/ML methods in the life sciences are published far faster than anyone can assess them. A reader — or a reviewer — often cannot tell from a paper how the data were split, whether the test set was independent, what the baseline was, or whether the model can be run at all. That is not a peer-review failure so much as a reporting failure: the information was never asked for in a structured way.
The DOME recommendations are the community’s answer to that, and this tutorial is about putting them into practice — not just reading them. We will look at the recommendations themselves, at the registry that makes DOME reports findable and reusable, at the tooling that makes producing one tractable at scale, at the broader Open and Sustainable AI picture, and then you will produce a report yourself by whichever submission route suits you.
Run sheet — the 90 minutes
| Time | Activity | Description |
|---|---|---|
| 14:00–14:10 | Introduction | Attendee background Learning outcomes Overview |
| 14:10–14:40 | Slides | AI/ML & publishing issues DOME Recommendations & Registry OSAI |
| 14:40–14:50 | Break | Brain rest |
| 14:50–15:25 | Hands-on tutorial exercise | AI/ML method reporting activity with: DOME Wizard DOME Copilot DOME Agent Skill |
| 15:25–15:30 | Wrap-up & discussion | Final Qs |
About MAQC 2026
This tutorial is part of the MAQC Society 2026 Annual Meeting, “Precision in a Complex World”, held 22–26 September 2026 at LIGHT, University of Brescia, Italy.
Credits, citation and licence
This tutorial draws on work by the ELIXIR Machine Learning Focus Group and its successor the AI Ecosystem Focus Group, BioComputingUP at the University of Padua, and the many contributors to the DOME and OSAI initiatives.
How to cite this tutorial
Farrell, G. (2026). Publishing Reusable AI/ML Models in the Life Sciences (MAQC 2026 Technical Tutorial P06). Zenodo. https://doi.org/10.5281/zenodo.22900624
Please also cite the underlying works listed in References.
Template attribution
This site is built with the ELIXIR Training Lesson Template by van Geest G, Kronander E, Romero Herrera JA, Žlender N, ELIXIR Training Coordination Team & Cardona A (2023), DOI 10.5281/zenodo.7913092, CC BY-SA 4.0. Content has been replaced with the MAQC 2026 tutorial material. Because the template is ShareAlike, this tutorial is also licensed CC BY-SA 4.0.