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

Area of Science:

• Biomedical Informatics
• Computational Biology
• Drug Discovery

Background:

• Traditional drug discovery faces low efficiency and high attrition due to disease complexity.
• Omics technologies (genomics, transcriptomics, proteomics) offer systems-level insights but present data challenges like high dimensionality and noise.
• Large language models (LLMs) show promise in addressing these omics data challenges by capturing intricate patterns and inferring information.

Purpose of the Study:

• To present a framework for integrating LLMs with omics data in drug discovery.
• To analyze LLM capabilities in handling omics data challenges.
• To detail LLM applications for uncovering disease mechanisms, identifying drug targets, predicting drug response, and simulating cellular behavior.

Main Methods:

• Analyzing LLM architectures and learning paradigms for omics data.
• Detailing LLM applications across key drug discovery stages.
• Discussing LLM-driven insights for pathway-targeted drug development.

Main Results:

• LLMs can effectively handle the complexity and noise inherent in omics datasets.
• LLM applications span disease mechanism elucidation, target identification, and drug response prediction.
• Omics-integrated LLMs facilitate a shift towards pathway-centric drug development strategies.

Conclusions:

• LLMs offer a powerful approach to overcome limitations in traditional drug discovery by leveraging omics data.
• This framework provides guidance for utilizing LLMs in omics-driven drug discovery and development.
• Integrating LLMs with omics data enables more effective and biologically grounded therapeutic strategies.