Filling the gaps: multi-experiment de novo tag integration for confident top-down protein characterization in OmniScape

Importance of the topic

The ability to perform reliable de novo interpretation of top-down tandem mass spectra is a key enabler for direct proteoform characterization, discovery of sequence variants and localization of post-translational modifications without prior database knowledge. Challenges such as complex spectral patterns, varied fragmentation behavior, incomplete sequence coverage and instrument-specific artifacts reduce confidence in conventional top-down identifications. Methods that integrate multiple complementary fragmentation modes and multiple experiments to build longer, higher-confidence sequence tags directly address these limitations and expand the practical reach of top-down proteomics for research and advanced QC workflows.

Objectives and overview of the study

This work presents a new de novo sequencing algorithm implemented in the OmniScape software that (1) generates high-confidence sequence tags by integrating MSn data from multiple experiments and fragmentation modes recorded on the timsOmni platform, (2) constructs composite spectral graphs using deconvoluted isotopic components, (3) scores tags with composite metrics, and (4) submits tags to a local alignment engine (LYRA) that accounts for isobaric segments to improve top-down homology searches. The study benchmarks performance on annotated spectra of bovine carbonic anhydrase II (CA II) and histone H3.1 collected with multimodal fragmentation strategies.

Methodology

The computational workflow and experimental design include the following elements:

• Data acquisition: Direct infusion top-down MS/MS on a timsOmni mass spectrometer using complementary fragmentation methods (examples in the study: ECD, CID, EID, ECciD and variations labeled CCID/RCID) across multiple precursor charge states.
• Preprocessing and deconvolution: An OmniWaveTM deconvolution algorithm implemented in C++ extracts averagine-based isotopic distributions from MSn spectra to provide monoisotopic components for downstream graph construction.
• Spectral graph construction: Deconvoluted fragment masses are used to build spectral graphs that represent observed mass differences and candidate residue links; composite graphs are generated by integrating graphs from multiple experiments and charge states.
• Tag scoring and selection: Sequence tags are evaluated using composite scoring metrics that account for both fragmentation feature quality (signal, continuity) and statistical significance.
• Local alignment and homology search: The selected tags are submitted to LYRA, a local-alignment search tailored for top-down data. LYRA treats isobaric sequence segments (e.g., I/L and some ambiguous substitutions) as valid matches rather than mismatches, improving sensitivity over standard MS-BLAST-like scoring.
• Performance assessment: The algorithm was tested on annotated MS/MS spectra of CA II and histone H3.1, comparing individual-experiment tags versus multi-experiment integrated tags for sequence coverage (SC %) and alignment bit scores.

Used instrumentation

• timsOmniTM mass spectrometer (Bruker) used for direct infusion top-down experiments.
• Multimodal fragmentation enabled on the platform: electron-based fragmentation (ECD), collision-induced dissociation (CID), electron-induced dissociation (EID), and hybrid variants (ECciD, CCID, RCID) across multiple precursor charge states.
• OmniScapeTM software suite incorporating the OmniWaveTM deconvolution algorithm and the LYRA local-alignment module for tag-based homology searching.

Main results and discussion

Key experimental outcomes demonstrate the value of multi-experiment integration:

• Composite sequence tags produced by merging spectral graphs from different precursor charge states and fragmentation modes substantially increased sequence coverage compared with tags from single experiments. Example: bovine CA II individual charge-state SC values were ~22.3% (z31), 32.3% (z34) and 28.1% (z37), while the combined z31+z34+z37 analysis reached ~44.6% SC.
• For histone H3.1, multimodal integration (EID + ECD + CID plus multiple charge states) raised SC to ~66.2%, compared with much lower coverage from single-method experiments (examples: ECD ~19.9%, CID ~14.0%, EID ~39.7%).
• Integration also improved homology search metrics. In one region shown, combining tags from multiple experiments increased the LYRA alignment bit score by approximately 57% (example from ~35 to ~55), indicating a markedly reduced chance of random matches and higher-confidence identifications.
• LYRA’s explicit treatment of isobaric segments (shown visually in alignments) recovered matches that conventional MS-BLAST-style scoring would mark as gaps or mismatches, yielding longer alignments and higher bit scores.
• Algorithmic implementation is computationally efficient: multi-experiment tag generation and alignment examples reported run-times on the order of seconds (~10 s in shown cases), supporting practical application in analysis workflows.

Benefits and practical applications

The integrated multi-experiment de novo tag approach delivers several practical advantages:

• Improved sequence coverage and longer tags reduce ambiguity in proteoform assignment and increase confidence in localizing modifications and sequence variants.
• Combining complementary fragmentation modalities compensates for method-specific blind spots and yields more comprehensive fragmentation maps, particularly valuable for challenging proteins like histones.
• LYRA’s isobar-aware alignments enhance sensitivity in homology searches, enabling better database matching in cases with isobaric residues or ambiguous segments.
• Fast execution and integration into the OmniScape environment make the workflow suitable for routine top-down characterizations in research and advanced QA/QC settings.

Future trends and potential uses

Anticipated developments and applications emerging from this work include:

• Tighter integration of de novo tag generation with machine-learning models for improved scoring, PTM localization and error modeling.
• Extension to higher-order MSn experiments and automated selection of complementary fragmentation strategies to maximize unique fragment yield.
• Broader adoption in proteoform-resolved biomarker discovery, structural proteomics and in workflows that require confident identification of sequence variants without complete database matches.
• Refinements to handle larger proteoforms and more complex PTM patterns, and to improve throughput for LC-MS coupling in continuous acquisition modes.
• Standardization of isobaric-aware scoring across tools to harmonize top-down homology reporting and reduce false negatives arising from conventional mismatch penalties.

Conclusions

Integrating multiple experiments and complementary fragmentation modes into a unified de novo tag generation workflow substantially improves top-down proteoform characterization. The OmniScape implementation—combining OmniWave deconvolution, spectral-graph integration and LYRA local alignments that accept isobaric matches—produces longer, higher-confidence tags, improves sequence coverage and increases alignment scores versus single-experiment approaches. These advances make de novo top-down sequencing more robust and practical for routine applications in research and specialized analytical laboratories.

Reference

• ASMS 2026, THP649: Filling the gaps: multi-experiment de novo tag integration for confident top-down protein characterization in OmniScape. Authors: George Alevizos, Mariangela Kosmopoulou, Georgia Orfanoudaki, Detlev Suckau, Athanasios Smyrnakis, Liliane Soares, Dimitris Papanastasiou. Bruker-related work (2026).