Application of LLMs/Transformer-Based Models for Metabolite Annotation in Metabolomics

LC-MS untargeted metabolomics is widely used, but metabolite annotation remains a key challenge. Transformer-based large language models (LLMs), trained on MS and chemical data, show great potential in tasks like retention time prediction and structure annotation. They also enable integration of multi-omics data. This review discusses their applications and future potential in improving annotation accuracy and advancing precision medicine.