Omics-based large language models: A new engine for drug discovery innovation

Traditional drug discovery suffers from low efficiency and high attrition rates, largely due to the complexity and heterogeneity of human diseases. Omics technologies offer a systems-level perspective for uncovering disease mechanisms and identifying therapeutic targets, but present challenges such as high dimensionality, noise, and heterogeneity. Large language models (LLMs), originally developed for natural language processing, are emerging as powerful tools to address these issues by capturing complex patterns and inferring missing information from large, noisy datasets. We present a three-part framework: (1) Analyzing how LLM architectures and learning paradigms handle challenges specific to genomics, transcriptomics, and proteomics data; (2) Detailing LLM applications in key areas: uncovering disease mechanisms, identifying drug targets, predicting drug response, and simulating cellular behavior; (3) Discussing how insights from omics-integrated LLMs can inform the development of drugs targeting specific pathways, moving beyond single targets towards strategies grounded in underlying disease biology. This framework provides both conceptual insights and practical guidance for leveraging LLMs in omics-driven drug discovery and development.