Omics Data Discovery Agents: Agent-Supported Retrieval, Reanalysis, and Synthesis of Published Omics Data

Area of Science:

• Bioinformatics
• Computational Biology
• Data Science

Background:

• Vast amounts of omics data are published but remain computationally inaccessible.
• Essential information for data reuse is often dispersed across publications and supplementary materials.

Purpose of the Study:

• To develop an agentic framework for automated retrieval, reanalysis, and synthesis of published omics data.
• To enhance the computational reuse of biomedical omics datasets.

Main Methods:

• Utilized large language model (LLM) agents with tools for data fetching, metadata extraction, and pipeline execution.
• Employed model context protocol (MCP) servers to expose containerized analysis tools.
• Performed end-to-end reanalyses of proteomics and bulk RNA-seq data.

Main Results:

• Successfully cataloged dataset references across thousands of articles.
• Achieved high correlation (0.85-0.997) and concordant differentially expressed features (Spearman 0.88-0.91) in five reanalyses.
• Demonstrated the feasibility of automated omics data reanalysis with minimal residual differences.

Conclusions:

• The agentic framework significantly improves the accessibility and reusability of published omics data.
• Automated reanalysis pipelines can reproduce key findings, facilitating large-scale data synthesis.
• This approach paves the way for more integrated and efficient biomedical data research.