Against published models on 70 species (4.9 M spectra), all decoding greedily, peptide precision: dIon 0.602, InstaNovo 0.570, Casanovo 0.492. dIon is highest in each of the 70 species. 6/9

Show Post

Here: huggingface.co/datasets/alf...

Show Post

Self-supervised pretraining can improve de novo peptide sequencing. To our knowledge, our new model dIon is the first to show it. Preprint with Alfred Nilsson, Joel Lapin, Samuel Payne and Mathias Wilhelm. doi.org/10.48550/arX... 1/9

Show Post

6/6 yHydra is a foundation model trained on nearly 20 million peptide-spectrum matches from 67 public repositories, enabling a unified representation of spectra and peptides. It is made for next-gen proteomics, from de novo sequencing to multi-omics integration, Metaproteomics and proteogenomics.

Show Post

After years of development, collaboration, and hard work, it's finally here! Now published in Bioinformatics: Foundation model enables interpretable open and error-tolerant searching for mass spectrometry-based proteomics academic.oup.com/bioinformati... #TeamMassSpec #DeepLearning #AI #Proteomics

Show Post