Machine learning SIG at the Galaxy Community Conference 2026

At the Galaxy Community Conference 2026 in Clermont-Ferrand, France, 22–26 June, the Galaxy Machine Learning community showcased a strong and connected vision for making AI/ML more usable, reproducible, and FAIR in life-science data analysis. The following sections highlight multiple contributions from the Galaxy ML community, including a poster, an ELIXIR BioHackathon talk, a training session, and CoFest work.

Poster

The ML in Galaxy poster highlighted the expanding Galaxy AI/ML ecosystem, including RAG workflows, Flexynesis for AI-powered multi-omics integration, TabPFN for fast tabular prediction, GPU-enabled JupyterLab for scalable model development, Galaxy-ML tools powered by Scikit-learn, XGBoost and TensorFlow, text-segmentation workflows using DocLayout-YOLO and Hugging Face’s Galaxy integration, and ongoing work on inference and fine-tuning of biological foundation models such as DNABERT-2, ProtBERT, ProtT5 and ESM.

ELIXIR BioHackathon talk on “Bridging DOME, BioAIrepo, BioModels and Galaxy”

The project Project 13: ELIXIR AI Ecosystem Integration: Bridging DOME, BioAIrepo and Galaxy was accepted by ELIXIR for the upcoming BioHackathon 2026. I presented the project (from Galaxy side) in the ELIXIR BioHackathon talk at GCC 2026. The talk highlighted the need for a more connected ecosystem of AI/ML tools and resources in life sciences, The ELIXIR talk presented potential ideas for bridging DOME, BioAIrepo, and Galaxy to support the full ML/AI project lifecycle: discovering models and metadata, training or fine-tuning models in Galaxy, evaluating them on unseen data, and sharing trained models and metadata back to public repositories such as BioAIrepo, BioModels, Europe PMC, bio.tools, Zenodo and Dataverse. Some of the research data repositories have already been integrated with Galaxy, and the talk highlighted the potential work to integrate BioAIrepo and others with Galaxy.

ML training: Image classification with GLEAM Image Learner

Michelle and I conducted the ML training session using GLEAM Image Learner tutorial, demonstrating practical image-learning tools in Galaxy for biomedical applications including balanced train/validation/test splitting, use of a pretrained CaFormer S18 384 backbone, and evaluation with accuracy, weighted precision, recall and F1 for imbalanced medical imaging datasets.

CoFest

During CoFest, I joined “Driving Galaxy using agents” project created by Dannon where I created a pull request (PR) to collect reference URLs in a response generated by LLM. The PR validates all the generated links and adds a references button shows all links returned in AI responses, improving transparency and trust for the LLM generated responses.

Acknowledgements

Special thanks to the organizers of GCC 2026 in Clermont-Ferrand, France for providing a platform to showcase our work and foster collaboration within the Galaxy community. I would like to thank the Galaxy ML community (Michelle, Paulo, Dannon, Jeremy and others) for their contributions to the poster, ELIXIR BioHackathon talk, training session, and CoFest work. I would also like to thank the ELIXIR BioHackathon organizers for accepting our project and Gavin and Eliot for their work on the proposal. Additionally, I would like to acknowledge the support from the University of Freiburg for funding my participation in the conference.