Precise and accurate real-time de novo sequencing of timsTOF data with the Novor algorithm on the Bruker ProteoScape™ platform

Importance of the Topic

Real-time de novo peptide sequencing is critical for applications where database references may be incomplete or unavailable, such as immunopeptidomics and metaproteomics. Integrating high-speed algorithms directly on acquisition platforms enhances throughput and enables on-the-fly decision making.

Objectives and Study Overview

This work describes the integration of the Novor de novo sequencing algorithm, optimized for timsTOF PASEF data, into the Bruker ProteoScape (BPS) platform. The aims are to improve sequencing accuracy, maintain speed compatible with real-time acquisition, and demonstrate performance across diverse samples, including mixed-species and immunopeptidomic datasets.

Methodology and Instrumentation

BPS Novor was trained on over 1.78 million peptide spectrum matches (PSMs) derived from ProLuCID-GPU ground truth filtered at 1% PSM false discovery rate. Training data included deeply fractionated and non-tryptic digests (GluC, Pepsin, Elastase, Chymotrypsin, Trypsin) of K562 lysates. Offline comparisons employed MGF files to eliminate preprocessing biases. Amino acid and peptide precision and recall metrics were calculated following established protocols.

• Mass spectrometry platform: Bruker timsTOF with Parallel Accumulation–Serial Fragmentation (PASEF)
• Software environment: Bruker ProteoScape real-time processing framework
• De novo engine: Rapid Novor Inc. algorithm optimized for timsTOF data ("BPS Novor")

Main Results and Discussion

Performance evaluation on a mixed-species (Human, Yeast, E. coli) tryptic digest showed that BPS Novor achieved on average a 5% increase in correct amino acid identification versus Software A and an 11% improvement over pre-trained Novor. Precision-recall curves at both amino acid and peptide levels were consistently higher for BPS Novor across all species segments. A representative spectrum analysis highlighted BPS Novor correctly assigning 12 of 14 amino acids compared to only 3 of 14 by Software A. Processing benchmarks demonstrated that BPS Novor operates 20× faster than competing tools and can keep pace with PASEF acquisition in real time.

Benefits and Practical Applications

By embedding a fast, accurate de novo sequencing engine into the ProteoScape platform, laboratories gain Run-&-Done capabilities for complex proteomics applications where database searches are insufficient. Key advantages include:

• Enhanced sensitivity for low-abundance peptides in immunopeptidomics
• Broad applicability to non-tryptic and metaproteomic samples
• Substantial reductions in data processing time
• On-the-fly sequencing decisions to guide acquisition strategies

Future Trends and Potential Applications

Ongoing developments may leverage machine learning to further refine scoring functions and extend de novo accuracy for post-translational modifications. Integration with cloud-based workflows and expansion to other ion mobility platforms will broaden adoption. Emerging applications include real-time discovery of neoantigens and environmental proteome profiling.

Conclusion

The integration of a timsTOF-optimized Novor algorithm into Bruker ProteoScape delivers a high-throughput, accurate de novo sequencing solution. BPS Novor outperforms existing tools in both identification metrics and processing speed, enabling novel real-time proteomics applications.

Reference

1. Xu X, et al. ProLuCID-GPU database search approach. 2015.
2. Tabb DL, et al. DTASelect: PSM FDR filtering. 2002.
3. Prianichnikov N, et al. Mixed-species timsTOF dataset PXD014777. 2020.
4. Feola SP, et al. Immunopeptidomic MHC class I dataset analysis. 2021.