Cross-linked mass spectrometry (XL-MS) plays an increasing role in the study of protein structure and protein-protein interactions, but the development of its data analysis methods is currently lagging that of conventional proteomics, especially in terms of artificial intelligence-based analysis and data-independent acquisition (DIA) methods. Here, we constructed Deep4D-XL, a deep learning tool capable of accurately predicting crosslinked peptide 4D information. Using Deep4D-XL as the core, we developed XL-MSDigger, a deep-learning-based pipeline for comprehensive data analysis of crosslinked mass spectrometry, which can realize deep analysis of cross-linked mass spectrometry DDA and DIA data. In the DDA data analysis of Hela crosslinked samples, we improved the identification of inter-protein crosslinked peptides by 62% through the rescoring method of XL-MSDigger. In DIA analysis, we realized the first FDR evaluation of DIA XL-MS and DIA XL-MS analysis based on predicted spectral libraries. In the analysis of heat shock protein PPI networks, DIA XL-MS based on predicted spectral libraries exhibited higher identification coverage than the DDA method. We anticipate that XL-MSDigger will play a pivotal role in advancing future XL-MS research.