Latent imputation
2 methods · 2025–2026
Latent imputation: Treats missing fragmentation as something to reconstruct rather than tolerate: latent representations of the theoretical peaks a peptide should have produced are imputed before the sequence is predicted.
Treats missing fragmentation as something to reconstruct rather than tolerate: latent representations of the theoretical peaks a peptide should have produced are imputed before the sequence is predicted.
The earliest of its 2 methods is LIPNovo (2025).
| Methods | 2 |
| Papers describing them | 2 |
| Authors | 7 |
| Active | 2025-05-23 to 2026-06-06 |
| Deep learning | 2 of 2 |
| Kinds | algorithm (2) |
| Acquisition | DDA (2) |
Methods (2)
Oldest first, by the paper that describes each one.
- LIPNovo (2025): Latent imputation
- LIPNovo+ (2026): Self-reflective extension of LIPNovo that finds the fragmentation sites where latent imputation is unreliable and refocuses training on them through a reflection-guided curriculum, so the model is not limited to the missingness patterns seen during training.
Papers describing them (2)
- Latent Imputation before Prediction: A New Computational Paradigm for De Novo Peptide Sequencing (2025, ICML 2025, preprint)
- LIPNovo+: Self-Reflective Latent Imputation for Robust De Novo Peptide Sequencing (2026, SSRN Electronic Journal, preprint)