Research on LC-MS Peptide Feature Detection Models Based on Deep Learning

In recent years, liquid chromatography-mass spectrometry (LC-MS) based quantitative proteomics has demonstrated significant potential. It enables the quantification of relative protein abundance in diverse biological fluid samples, establishing itself as a crucial technique for disease biomarker discovery. One of the critical steps in identifying disease biomarkers is the detection of peptide features, along with their corresponding charges and intensities, within LC-MS spectra. However, the existing heuristic algorithms are susceptible to inaccuracies in parameters and potential human errors. Thus, effectively leveraging deep learning for peptide feature detection has become a pressing concern. This study focuses on analyzing models such as DeepIso, PointIso, and MSTracer for addressing the LC-MS peptide feature detection problem through deep learning approaches. It offers a variety of guidance and methodologies for peptide feature detection. This study focuses on a comparative evaluation and synthesis of the detection accuracy, advantages, and limitations of existing models in LC-MS data. This facilitates further investigation into peptide feature detection issues and provides guidance for future research directions.