Spectra Fragment-Ion and Amino Acid Probability Prediction for Peptide Sequencing
thesis · 2026
| Date | 2026-08-18 |
| Type | thesis |
| Publisher | MSc thesis |
| Contribution | adjacent |
| Supervisor | Kaizhong Zhang |
| Link | https://hdl.handle.net/20.500.14721/40424 |
Abstract
Reliable tandem mass-spectral analysis requires distinguishing sequence-informative fragment evidence from noise and incomplete observations. This problem is important in peptide-spectrum interpretation, including database-assisted analysis and de novo peptide sequencing, because an observed peak is not necessarily a sequence-informative fragment, whereas an unobserved fragment may still be valid. This thesis formulates two local probability-prediction tasks. Fragment-Ion Probability estimates whether a peak-supported mass position represents a sequence-informative fragment ion, and Amino Acid Probability estimates whether two mass-consistent positions are adjacent fragments connected by a candidate residue. CNN and encoder-only Transformer models are developed for both tasks. In the controlled downstream experiment, using the evaluated probability models achieve higher Amino Acid Recall, Amino Acid Precision, and Peptide Recall, showing that the predicted probability provides useful complementary spectral evidence.
Methods and tools
- Casanovo: First Transformer
- DpNovo: Transformer + dynamic programming
- Fragment-ion and amino-acid probability models: Two local probability-prediction tasks for tandem mass spectra, aimed at telling sequence-informative fragment evidence from noise and from merely unobserved fragments. Fragment-Ion Probability scores whether a peak-supported mass position is a sequence-informative fragment ion; Amino Acid Probability scores whether two mass-consistent positions are adjacent fragments joined by a candidate residue. Encoder-only Transformers outperform CNN baselines on both, and feeding the probabilities to a separate sequencer raises amino-acid recall, amino-acid precision and peptide recall.