Following the adoption of the DOME machine learning recommendations by GigaScience Press, publishers of machine learning life science research are being encouraged to follow suit.
The DOME (data, optimisation, model and evaluation) recommendations were developed by the ELIXIR Machine Learning Focus Group and described in a 2021 paper, since viewed over 23,000 times. Last year, the DOME registry was launched, providing a curated database of DOME annotations across a number of supervised machine learning models in biology, encompassing both published and unpublished studies.
After a successful trial, GigaScience has adopted the DOME recommendations as part of their peer review and publication process for machine learning papers. Researchers are helped to generate DOME annotations using the DOME Wizard, which also submits the annotations and shares them with reviewers. A link to DOME annotation is included in the dataset accompanying a manuscript, and deposited in the DOME registry.
Chris Armit, a Data Scientist at GigaScience, says “The DOME annotations are a great asset to peer review, providing the necessary high-level overview to properly understand a machine learning study. We recommend that other journals follow our example in encouraging DOME annotations to be submitted early in the publication process and prior to peer-review”.
The DOME Registry is designed to fit into a publisher’s workflow to enable an objective and transparent assessment of the methods used in a machine learning manuscript. The annotations are particularly helpful for reviewers who may be experts in areas of life science or bioinformatics but struggle to identify opaque, poorly described, or even deceptive machine learning methods.
By supporting journals to publish open and high-quality machine learning papers, the DOME standards gives the broader academic community access to more reliable data to enable better research and innovation.
DOME is led by Silvio Tosatto (University of Padua), ELIXIR Italy, and Fotis Psomopoulos (Centre for Research and Technology Hellas - CERTH), ELIXIR Greece, who can be contacted at contact@dome-ml.org.