Large language models for biological sequence analysis in infectious disease research.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Large language models (LLMs) show significant promise for analyzing complex biological data.
• Their application in infectious disease research is rapidly expanding, offering new analytical capabilities.

Purpose of the Study:

• To review the primary types of biological LLMs: protein, genomic, and multimodal models.
• To highlight the architectures and diverse applications of these models in infectious disease research.
• To discuss current challenges and future directions for LLM implementation.

Main Methods:

• Review of existing literature on biological LLMs.
• Categorization of LLMs into protein, genomic, and multimodal types.
• Analysis of LLM applications in pathogen identification, evolutionary surveillance, host-pathogen prediction, and therapeutic development.

Main Results:

• LLMs enable large-scale interpretation of genomic and proteomic data.
• Key applications include enhanced pathogen identification and accelerated therapeutic development.
• Identified challenges include data quality, model interpretability, and biosafety.

Conclusions:

• Biological LLMs are transformative tools for infectious disease research.
• Addressing challenges in data quality, interpretability, and safety is crucial for responsible deployment.
• Understanding LLM potential and limitations is key to advancing infectious disease science.