Proteome Discoverer 3.0 software with the CHIMERYS intelligent search algorithm

Importance of the Topic

Modern proteomics relies on high‐resolution tandem mass spectrometry to identify and quantify thousands of peptides in complex mixtures. Classical database search algorithms often fail to interpret chimeric spectra that contain fragments from multiple co‐isolated peptides. This limitation reduces the number of peptide‐spectrum matches (PSMs), hinders protein coverage and slows biological discovery.

Study Objectives and Overview

This work evaluates the CHIMERYS intelligent search algorithm integrated into Thermo Scientific Proteome Discoverer 3.0. The goal is to assess how an artificial intelligence–driven approach can improve PSM identification rates, peptide and protein coverage, and support specialized applications such as HLA immunopeptidomics, in comparison to prior versions and the SEQUEST HT algorithm with INFERYS rescoring.

Methodology

The CHIMERYS algorithm applies deep learning to deconvolute chimeric tandem mass spectra and assign multiple PSMs per MS/MS scan. Data‐dependent acquisitions were performed on Thermo Scientific Orbitrap mass spectrometers across a range of protein loads (10–500 ng) and chromatographic run times (30–120 min). Identifications using CHIMERYS were compared to PD 2.3–2.5 and SEQUEST HT with INFERYS rescoring, evaluating metrics such as PSMs per spectrum, unique peptides and proteins. A publicly available HLA Class I data set from a patient‐derived melanoma line was also reanalyzed to benchmark immunopeptidomic performance.

Used Instrumentation

• Thermo Scientific Orbitrap mass spectrometers
• Proteome Discoverer 3.0 software
• CHIMERYS intelligent search algorithm (MSAID GmbH)
• INFERYS 2.0 rescoring
• SEQUEST HT search engine

Main Results and Discussion

CHIMERYS delivers a marked increase in PSMs per spectrum, with many spectra yielding three or more matches compared to one in prior versions. Unique peptide identifications increased by up to 128% and unique proteins by up to 76% under optimized loading and gradient conditions. Improvements are especially pronounced for shorter gradients and higher sample loads, enabling faster throughput. In HLA immunopeptidomics, INFERYS rescoring within PD 3.0 produced a 55% rise in peptide identifications at 1% false discovery rate, demonstrating enhanced coverage of complex peptide repertoires.

Benefits and Practical Applications

• Enhanced depth in proteome profiling and quantitation accuracy
• More efficient sample and instrument utilization through shorter runs
• Improved pathway coverage for biological insights
• Superior support for immunopeptidomic workflows at stringent FDRs

Future Trends and Opportunities

The integration of AI into proteomics datasets will continue to expand with possibilities including real‐time spectrum deconvolution, adaptation to data‐independent acquisition approaches, single‐cell proteomics and clinical diagnostic pipelines. Continuous algorithm refinement and larger training sets will further drive sensitivity and specificity improvements.

Conclusion

The AI‐driven CHIMERYS search algorithm in Proteome Discoverer 3.0 overcomes limitations of traditional search strategies by resolving chimeric spectra and maximizing peptide and protein identifications. This advancement empowers researchers with deeper proteomic insights, faster workflows and enhanced support for specialized applications such as immunopeptidomics.

References

• Chong C et al (2020) Nature Communications