Enhancing phenotype recognition in clinical notes using large language models: PhenoBCBERT and PhenoGPT

• DOI: 10.1016/j.patter.2023.100887
• Publication Date: 2023-12-05T16:35:19Z
• Authors: Jingye Yang; Cong Liu; Wendy Deng; Da Wu; Chunhua Weng; Yunyun Zhou; Kai Wang
• Keywords: Leverage (statistics); Scope (computer science)

To enhance phenotype recognition in clinical notes of genetic diseases, we developed two models - PhenoBCBERT and PhenoGPT for expanding the vocabularies Human Phenotype Ontology (HPO) terms. While HPO offers a standardized vocabulary phenotypes, existing tools often fail to capture full scope due limitations from traditional heuristic or rule-based approaches. Our leverage large language (LLMs) automate detection terms, including those not current HPO. We compared these PhenoTagger, another tool, found that our identify wider range concepts, previously uncharacterized ones. also showed strong performance case studies on biomedical literature. evaluated strengths weaknesses BERT-based GPT-based aspects such as architecture accuracy. Overall, automated texts, improving downstream analyses human diseases.