pGlycoQuant with a deep residual network for quantitative glycoproteomics at intact glycopeptide level
Siyuan Kong, Pengyun Gong, Wen‐Feng Zeng, Biyun Jiang, Xinhang Hou, Yang Zhang, Huanhuan Zhao, Mingqi Liu, Guoquan Yan, Xinwen Zhou, Xihua Qiao, Mengxi Wu, Pengyuan Yang, Chao Liu, Weiqian Cao
Abstract
Large-scale intact glycopeptide identification has been advanced by software tools. However, tools for quantitative analysis remain lagging behind, which hinders exploring the differential site-specific glycosylation. Here, we report pGlycoQuant, a generic tool for both primary and tandem mass spectrometry-based intact glycopeptide quantitation. pGlycoQuant advances in glycopeptide matching through applying a deep learning model that reduces missing values by 19-89% compared with Byologic, MSFragger-Glyco, Skyline, and Proteome Discoverer, as well as a Match In Run algorithm for more glycopeptide coverage, greatly expanding the quantitative function of several widely used search engines, including pGlyco 2.0, pGlyco3, Byonic and MSFragger-Glyco. Further application of pGlycoQuant to the N-glycoproteomic study in three different metastatic HCC cell lines quantifies 6435 intact N-glycopeptides and, together with in vitro molecular biology experiments, illustrates site 979-core fucosylation of L1CAM as a potential regulator of HCC metastasis. We expected further applications of the freely available pGlycoQuant in glycoproteomic studies.