Omics-scale decoding of protein N-glycosylation using protein language model embeddings

MaskNGly is a zero-shot framework that identifies protein N-glycosylation sites from protein language model embeddings with performance comparable to state-of-the-art supervised predictors.

• Why it matters: Interpretability and scalability are limited because pLM embeddings are a black box and supervised glycosylation predictors require downstream training.
• What they did: The method decodes semantic features from pLM embeddings using coevolutionary analysis, then performs site identification without supervised training.
• The result: Across proteomes, it links higher glycosylation propensities to canonical N-X-S/T motifs in more complex organisms, favors structured solvent-exposed regions, and finds that mutations in high-scoring motifs are more often disease-associated.

AI summary of the abstract · may contain errors · not medical advice · read the paper.