Large language models for biological sequence analysis in infectious disease research

• Large language models (LLMs) are transforming infectious disease research by analyzing complex biological sequences at an unprecedented scale. • Three main types of biological LLMs are used: protein language models, genomic language models, and multimodal models • LLMs have numerous applications, including pathogen identification, evolutionary surveillance, host-pathogen prediction, and therapeutic development. • Despite their potential, challenges remain in data quality, long-context processing, model interpretability, translation to practice, and biosafety. Large language models (LLMs) have emerged as transformative tools in infectious disease research, offering unprecedented capabilities in analyzing biological sequences. This review summarizes three primary types of biological LLMs, including protein language models, genomic language models, and multimodal models, highlighting their architectures and applications. These models are revolutionizing key areas such as pathogen identification, evolutionary surveillance, host-pathogen prediction, and therapeutic development by enabling the interpretation of complex genomic and proteomic data at an unparalleled scale. While recent advancements are remarkable, challenges persist in data quality, long-context processing, model interpretability, and biosafety considerations. Understanding the potential and limitations of LLMs is crucial for leveraging them effectively in infectious disease research while ensuring responsible development and deployment.