质谱从头测序算法的发展:从图论到深度学习OA

Mass spectrometry-based proteomics is a crucial technique for elucidating protein sequences and functions in life sciences.Traditional database search methods require high completeness of reference genomes,making them unsuitable for analyzing non-model organisms,mutated peptides,or samples with unknown post-translational modifications.In contrast,de novo sequencing algorithms can directly infer amino acid sequences from tandem mass spectrometry fragment data,offering broad applications in areas such as antibody proteomics and protein identification.Over the past three decades,de novo sequencing algorithms have evolved through stages including graph theory,dynamic programming,and machine learning,and have now fully entered the era of deep learning.Different de novo algorithms are suited to distinct fragmentation patterns and vary in performance regarding speed and cross-dataset analysis capabilities.This review summarizes the development history of mass spectrometry de novo sequencing algorithms,highlights representative approaches from four key phases,including graph theory,dynamic programming,machine learning,and deep learning,discusses current challenges,and outlines future research directions.