Optimizing peptide identification for mass spectrometry-based plasma proteomics

The discovery of plasma protein biomarkers is critical for early disease diagnosis. Mass spectrometry-based proteomics serves as a primary approach to identify and quantify proteins within biological materials such as blood plasma. The typical workflow involves digestion of proteins into smaller molecules called peptides. Subsequently, these peptides are separated into smaller fragments. Fragmentation of peptides is recorded as mass spectra. To identify these peptides, experimental spectra are matched against a reference spectral library. Consequently, selecting a representative mass spectrum for each peptide is essential. This thesis proposes a method to select the most appropriate fragmentation spectrum for each peptide from multiple associated observed spectra. The method leverages prototypical networks, utilizing a Transformer encoder as the embedding function to evaluate and select the best representative spectrum. Experimental results demonstrate that the performance of the proposed method is comparable to the established baseline, establishing a foundation for deep-learning-based consensus spectral library generation.

Từ khóa: proteomics, prototypical networks, transformer.

Nguyen Thai Huy, abstract of graduation thesis at the VNU University of Engineering and Technology, Hanoi 2026