LeGenD: high-throughput N-glycan profiling using explainable AI and lectin profiling

Abstract Glycosylation affects many vital functions in organisms. Thus, their measurement is critical from basic science to biotechnology, including biopharmaceutical development and clinical diagnostics. However, the throughput and cost of conventional glycan analysis can be challenging. Lectins offer an alternative approach for analyzing glycans, but they only provide glycan epitopes and not full glycan structure information. To overcome these limitations, we developed LeGenD, a lectin and AI-based approach, to predict dominant N-glycan structures and determine their relative abundance on purified proteins based on lectin-binding patterns. We trained the LeGenD model on 309 glycoprofiles from 10 recombinant proteins, produced in 30 glycoengineered CHO cell lines. Independent test data showed that the dominant glycosylation patterns in a given protein can be effectively determined. Further analysis using SHapley Additive exPlanations helped to identify critical lectins for glycoprofile predictions. Thus, our LeGenD approach presents an alternative platform for analyzing protein glycosylation and could complement the existing toolkits used to study glycosylation. Teaser LeGenD, a lectin profiling-based AI platform, predicts dominant N-glycan structures, providing an alternative method for protein glycosylation analysis.