GlycoGauntlet competition to annotate glycomics data

Welcome to the GlycoGauntlet competition!

Predict glycan structures from mass spectrometry glycomics data. Beat our deep learning model (CandyCrunch) or provide expert manual annotations. The goal of this competition is manifold: (i) assess the current state of algorithmic annotation solutions compared to human experts, (ii) benchmark overlap/concordance between different human annotators, (iii) formally quantify what is hard and what is easy in annotation, provide quantifiable estimates on uncertainties at a substructural level, etc etc

If you provide submissions for private files (one submission per file), you will be invited to be a co-author on the GlycoGauntlet paper (public file submissions optional but encouraged for an even better paper/comparison:-) ).

All files are negative ion mode, reduced animal glycans, run on a PGC column, and are provided as .mzML files. All files come from separate samples (not replicates) and represent a full LC-MS/MS run. Your job is to output glycan structures in any common notation (e.g., IUPAC-condensed, GlyTouCan IDs, etc), as well as where they're found (m/z + retention time).

There are two zipped folders in this dataset: public.zip (glycomics datasets where the answer is known, public, and stored in our competition repository, https://github.com/BojarLab/GlycoGauntlet) and private.zip (unpublished glycomics datasets where we have an expert annotation but it is not publicly available/known)

All datasets are either N- or O-glycomics and are labeled as such. Public file submissions are immediately scored and scores will be displayed on a public leaderboard. Private file submissions will also be scored but scores will be hidden until the end of the competition. You can submit as many attempts as you want

This competition is tentatively expected to conclude in October 2026 but we reserve the right to move this date as things progress