Gatto, LaurentGhislain, MichelDeflandre, GuillaumeGuillaumeDeflandre2025-05-142025-05-142025-05-142024https://hdl.handle.net/2078.2/37740Single-cell proteomics (SCP) has emerged as a powerful tool for elucidating cellular heterogeneity, offering opportunities beyond traditional bulk sample analysis. However, the application of current peptide identification algorithms crafted for bulk samples may lead to false discoveries in SCP. Challenges such as reduced peak counts, lower peak intensities, and degraded signal-to-noise ratios raise the question: do current peptide scoring methods in search engines adequately perform in the context of SCP? To address these limitations, we explore the effectiveness of database search engines and rescoring tools with the use of Bioconductor packages PSMatch and Spectra. Rescoring tools take profit of as many mass spectrometry-based features as possible, such as spectral characteristics and retention time models, which can be particularly relevant to mitigate the poor quality of SCP spectra. We used MS²Rescore to generate new features, Mokapot to rescore the SCP peptides as well as the above-mentioned packages to assess the efficiency of rescoring tools and potentially improve current scoring methods in the context of SCP. While the need for identification optimization at search engine level is not identified, optimization of rescoring tools is endorsed: our findings demonstrate a notable increase in confidently identified peptides upon rescoring. In addition, we suggest a four-step methodology to evaluate the usefulness of current and new potential features. Finally, our results shed light on the differences between bulk and single-cell samples whilst providing insights that can inform more accurate and reliable data interpretation in the context of SCP.ProteomicsSingle-cellMass spectrometryPeptide identificationSearch enginesRescoringtext::thesis::master thesisthesis:46486