CNN
3 methods · 2022–2024
CNN: Convolutional networks without a recurrent decoder, reading the spectrum as a signal over the m/z axis. Used where the output is a score or a local call rather than a sequence emitted left to right.
Convolutional networks without a recurrent decoder, reading the spectrum as a signal over the m/z axis. Used where the output is a score or a local call rather than a sequence emitted left to right.
The earliest of its 3 methods is PepNet (2022); 2 more have followed.
| Methods | 3 |
| Papers describing them | 5 |
| Authors | 21 |
| Active | 2022-02-09 to 2024-01-02 |
| Deep learning | 3 of 3 |
| Kinds | adjacent, algorithm, post-processor |
| Acquisition | DDA (3) |
Methods (3)
Oldest first, by the paper that describes each one.
- PepNet (2022): Temporal convolutional network
- Spectralis (2023): AA-gapped convolutional layer
- GlycanFinder (2023): Glycopeptide sequencing
How they score
2 of the 3 have been run on denovo_benchmarks, which ranks 17 tools over 84 datasets. The family’s best median rank is 8.
- Spectralis: median peptide-level average precision 0.725, median rank 8 of 17
- PepNet: median peptide-level average precision 0.604, median rank 11 of 17
Read these next to the rest of the field, not on their own: what the numbers mean.
Papers describing them (5)
- PepNet: A Fully Convolutional Neural Network for De novo Peptide Sequencing (2022, Research Square, preprint)
- Deep learning-driven fragment ion series classification enables highly precise and sensitive de novo peptide sequencing (2023, bioRxiv, preprint)
- Glycopeptide database search and de novo sequencing with PEAKS GlycanFinder enable highly sensitive glycoproteomics (2023, Nature Communications, peer-reviewed)
- Accurate de novo peptide sequencing using fully convolutional neural networks (2023, Nature Communications, peer-reviewed)
- Deep learning-driven fragment ion series classification enables highly precise and sensitive de novo peptide sequencing (2024, Nature Communications, peer-reviewed)