dIon: Fragmentation-Based Invariance for Self-Supervised Learning of Tandem Mass Spectra
Paper • 2610.06282 • Published
version int64 |
|---|
924 |
Datasets used in dIon. See the paper for original source accessions, citations, and experiment definitions.
pretraining/: bacterial SSL corpus and used 1%/10% metric-learning subsets.probes/end_aa/: shared online end-AA monitor.representation_benchmarks/: full-charge and separately constructed charge-2-4 retrieval/pair benchmarks.de_novo/: MSKB-final, DNL-v1, capped Kingdoms, mini-de-novo probe, and DIA pilot.auxiliary_spectrum_tasks/: SQA, HYE chimericity, and oxidized-methionine tasks.dataset_inventory.json lists the files included in this release.