![[2602.20209] Regressor-guided Diffusion Model for De Novo Peptide Sequencing with Explicit Mass Control](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.20209] Regressor-guided Diffusion Model for De Novo Peptide Sequencing with Explicit Mass Control
Summary
The paper presents DiffuNovo, a novel regressor-guided diffusion model for de novo peptide sequencing that incorporates explicit mass control, significantly improving accuracy over existing methods.
Why It Matters
This research addresses a critical limitation in peptide sequencing by ensuring predicted peptide masses align with experimental data. The introduction of DiffuNovo marks a significant advancement in the field, potentially enhancing protein identification and discovery in various biological applications.
Key Takeaways
- DiffuNovo integrates mass control during both training and inference stages.
- The model outperforms existing de novo peptide sequencing methods in accuracy.
- Utilizes a diffusion model architecture for enhanced controllability.
- Reduces mass error, producing more physically plausible peptide predictions.
- Source code is available, promoting further research and application.
Quantitative Biology > Quantitative Methods arXiv:2602.20209 (q-bio) [Submitted on 23 Feb 2026] Title:Regressor-guided Diffusion Model for De Novo Peptide Sequencing with Explicit Mass Control Authors:Shaorong Chen, Jingbo Zhou, Jun Xia View a PDF of the paper titled Regressor-guided Diffusion Model for De Novo Peptide Sequencing with Explicit Mass Control, by Shaorong Chen and 2 other authors View PDF HTML (experimental) Abstract:The discovery of novel proteins relies on sensitive protein identification, for which de novo peptide sequencing (DNPS) from mass spectra is a crucial approach. While deep learning has advanced DNPS, existing models inadequately enforce the fundamental mass consistency constraint, that a predicted peptide's mass must match the experimental measured precursor mass. Previous DNPS methods often treat this critical information as a simple input feature or use it in post-processing, leading to numerous implausible predictions that do not adhere to this fundamental physical property. To address this limitation, we introduce DiffuNovo, a novel regressor-guided diffusion model for de novo peptide sequencing that provides explicit peptide-level mass control. Our approach integrates the mass constraint at two critical stages: during training, a novel peptide-level mass loss guides model optimization, while at inference, regressor-based guidance from gradient-based updates in the latent space steers the generation to compel the predicted peptide adheres to ...
