Zero-shot de novo peptide sequencing with open posttranslational modification discovery

Area of Science:

• Proteomics
• Bioinformatics
• Computational Biology

Background:

• De novo peptide sequencing infers peptide sequences directly from mass spectrometry data, bypassing traditional database searches.
• Current deep learning methods for identifying posttranslational modifications (PTMs) necessitate labeled training data, limiting their scope.
• Discovering novel or unannotated PTMs remains a significant challenge in proteomics.

Purpose of the Study:

• To introduce RNovA, a transformer-based de novo sequencing algorithm designed for open posttranslational modification discovery.
• To enable the identification of PTMs in a zero-shot setting, without the need for retraining or a predefined list of modifications.
• To maintain high performance on standard benchmarks while expanding the capabilities of de novo sequencing.

Main Methods:

• Developed RNovA, a de novo sequencing algorithm utilizing rotary positional embeddings and a reinforcement-learning-style sequential decision framework.
• Implemented a transformer architecture enhanced with relative positional embeddings for improved sequence inference.
• Applied RNovA to analyze clinical samples and bacterial proteomes lacking reference databases.

Main Results:

• RNovA achieved state-of-the-art performance on standard de novo sequencing benchmarks.
• Successfully identified kynurenine-modified peptides in rheumatoid arthritis patient samples, validating findings with synthetic peptides.
• Detected an unannotated glutamic acid modification in the bacterial strain A1232E, demonstrating open PTM discovery capability.

Conclusions:

• RNovA facilitates open PTM discovery in a zero-shot manner, significantly broadening the scope of proteomic analysis.
• The algorithm enables the exploration of previously inaccessible proteomic regions, including peptides with unexpected or unannotated modifications.
• RNovA represents a significant advancement in de novo sequencing, offering powerful capabilities for uncovering novel biological insights.