GNN
3 methods · 2023
GNN: Graph neural networks over the spectrum graph: they learn the path-scoring function that dynamic programming had to hand-code, which is what lets a path cross a fragment that was never observed.
Graph neural networks over the spectrum graph: they learn the path-scoring function that dynamic programming had to hand-code, which is what lets a path cross a fragment that was never observed.
The earliest of its 3 methods is Hybrid CNN-GNN peptide ion encoder (2023); 2 more have followed.
| Methods | 3 |
| Papers describing them | 3 |
| Authors | 11 |
| Active | 2023-02-03 to 2023-10-19 |
| Deep learning | 3 of 3 |
| Kinds | algorithm (2), adjacent |
| Acquisition | DDA (3) |
Methods (3)
Oldest first, by the paper that describes each one.
- Hybrid CNN-GNN peptide ion encoder (2023): Convolutional plus graph neural network module that encodes fragment peaks to identify b- and y-ions, outperforming encoders taken from existing de novo sequencers under noise and missing peaks.
- Denovo-GCN (2023): GCN on spectrum graph
- GraphNovo (2023): Missing fragments
How they score
1 of the 3 has been run on denovo_benchmarks, which ranks 14 tools over 86 datasets. The family’s best median rank is 11.
- Denovo-GCN: median peptide-level average precision 0.559, median rank 11 of 14
Read these next to the rest of the field, not on their own: what the numbers mean.
Papers describing them (3)
- Application of a Novel Hybrid CNN-GNN for Peptide Ion Encoding (2023, Journal of Proteome Research, peer-reviewed)
- Denovo-GCN: De Novo Peptide Sequencing by Graph Convolutional Neural Networks (2023, Applied Sciences (MDPI), peer-reviewed)
- Mitigating the missing-fragmentation problem in de novo peptide sequencing with a two-stage graph-based deep learning model (2023, Nature Machine Intelligence, peer-reviewed)