Automated Glycan Composition Inference and Deconvolution via Multi-Modal Data Fusion and Graph Neural Networks: A Pipeline for Targeted Therapeutics Development

**Abstract:** Glycan analysis is crucial for understanding protein function and developing targeted therapies. Current methods are often laborious, error-prone, and limited in scalability. This paper introduces a novel pipeline, **GlycoInsight**, leveraging multi-modal data fusion and graph neural networks to automate glycan composition inference and deconvolution from mass spectrometry, nuclear magnetic resonance (NMR), and enzymatic digestion data. GlycoInsight achieves unprecedented accuracy and speed, offering a pathway for accelerating drug discovery by enabling rapid characterization of glycosylation patterns associated with disease.