IAN, an intelligent system for omics data analysis and discovery.

IAN is an R package addressing the challenge of integrating, analyzing, and interpreting gene expression data using an artificial intelligence (AI) system. It leverages popular pathway (KEGG, WikiPathways, Reactome, and GO) and regulatory datasets (ChEA), along with STRING for protein-protein interactions, to perform standard enrichment analysis. A multi-agent architecture uses individual enrichment results to generate summaries using large language models (LLMs). These summaries are contextually integrated and interpreted by the LLM, guided by engineered prompts and grounding instructions, to provide useful explanations, system overviews, key regulators, and data-driven insights. We demonstrate IAN's function on published human transcriptomic data, evaluate IAN's robustness to underlying LLM choice and report human expert evaluations of the clarity and relevance of IAN's outputs. The IAN package is available at https://github.com/NIH-NEI/IAN.