Beyond Best Guesses: A Multi-Level Confidence Assessment Framework for De Novo Peptide Sequencing Results

presentation · HUPO 2026 World Congress · 2026

presentation · HUPO 2026 World Congress · 2026. Cemil Can Saylan et al.
Date 2026-09-29
Type presentation
Venue HUPO 2026 World Congress
Publisher Human Proteome Organization
Contribution post-processor
Link https://tnuki.com/conferences/hupo-2026/technical-university-of-munich

Authors

  1. Cemil Can Saylan · Technical University of Munich
  2. Armin Suleymaniniya · Technical University of Munich
  3. Mathias Wilhelm · Technical University of Munich

Methods and tools

  • Peptide similarity alignment: Grades de novo sequencing errors against a reference peptide: Levenshtein-distance tiers plus a Needleman-Wunsch alignment aware of block shuffles and isobaric substitutions, so amino-acid accuracy can be split by error type. Paired with a one-class isolation-forest rescorer over Oktoberfest features, which filters identifications without FDR control.

Methods it uses

  • Casanovo: First Transformer
  • Prosit: Deep-learning (bidirectional RNN + attention) predictor of peptide fragment-ion intensities and retention times: widely used to validate and rescore de novo sequencing candidates against predicted spectra.

Seen in the charts

Back to the full map

Back to top