XL-MSDigger: a deep learning-based, versatile solution for cross-linking mass spectrometry

Cross-linking mass spectrometry (XL-MS) technology plays an important role in protein structure and protein-protein interaction (PPI) studies. However, there are serious shortcomings in the data processing and analysis methods of XL-MS that limit its ability to perform in-depth analysis. Here, we built a deep learning-based XL-MS versatile platform, XL-MSDigger, to address the bottlenecks of XL-MS technology. The foundation of this platform is a deep learning-based multidimensional information prediction tool, Deep4D-XL, for cross-linked peptides. This tool is capable of accurately predicting retention time, collisional cross-section, and fragment ion intensity information of cross-linked peptides. Through the multidimensional information prediction, we develop rescoring algorithms and workflows for DDA and DIA analysis data, which improve the coverage of DDA-based XL-MS identification, and realize the evaluation of false discovery rate and high-reliability identification of DIA-based XL-MS analysis. In addition, XL-MSDigger enables DIA-based XL-MS analysis using predicted spectral libraries, allowing improved detection of protein–protein interactions in targeted or moderate-scale interaction analyses. We believe that the XL-MSDigger platform can be widely used as a general solution to enhance the performance of XL-MS analysis. Cross-linking mass spectrometry is widely used to study protein structures and interactions but remains challenging to analyze comprehensively. Here, the authors develop a deep learning–based platform that enables deeper, systematic analysis of cross-linked peptides across DDA and DIA.