π-PrimeNovo: An Accurate and Efficient Non-Autoregressive Deep Learning Model for De Novo Peptide Sequencing
Lists
Only you see this
Abstract
Peptide sequencing via tandem mass spectrometry (MS/MS) is fundamental in proteomics research. In contrast to conventional database searching methods, deep learning models excel in de novo sequencing peptides absent from existing databases, thereby facilitating the identification of novel sequences. Current deep learning models for peptide sequencing predominantly use an autoregressive generation approach, where early errors can cascade, resulting in both undesired performance and low inference speed. To address this, we introduce $\pi$-PrimeNovo, a non-autoregressive Transformer-based deep learning peptide sequencing model. With the novel architecture design and the proposed CUDA-programming-aided dynamic decoding for precise mass controlling, $\pi$-PrimeNovo achieves significantly higher accuracy and up to 89x faster sequencing compared to the state-of-the-art methods, making it highly suitable for large scale sequencing tasks such as metaproteomic research. Moreover, $\pi$-PrimeNovo shows powerful capability in mining phosphopeptides and detecting low-abundance post-translational modifications (PTMs). Our work greatly advances peptide sequencing techniques and introduces a transformative computational model with broad implications for biological research.
BibTeX
@inproceedings{zhang2024primenovo,
author = {Xiang Zhang},
title = {{π-PrimeNovo: An Accurate and Efficient Non-Autoregressive Deep Learning Model for De Novo Peptide Sequencing}},
booktitle = {Nature Communications},
year = {2024},
url = {https://openreview.net/forum?id=AW0DaPH5aw},
}Loading related content...