IAN: An Intelligent System for Omics Data Analysis and Discovery

Area of Science:

• Bioinformatics
• Computational Biology
• Artificial Intelligence in Genomics

Background:

• High-throughput omics data analysis presents challenges in integration and interpretation.
• Existing methods may lack comprehensive analytical and interpretive capabilities.

Purpose of the Study:

• To introduce IAN, an R package designed for integrating, analyzing, and interpreting complex omics data.
• To leverage a multi-agent artificial intelligence (AI) system for enhanced biological insights.

Main Methods:

• Utilizes popular pathway and regulatory datasets (KEGG, WikiPathways, Reactome, GO, ChEA) and STRING for enrichment analysis.
• Employs a large language model (LLM) within a multi-agent architecture to summarize and interpret enrichment results.
• Applies carefully engineered prompts and grounding instructions for contextual integration and interpretation.

Main Results:

• IAN successfully reanalyzes published omics datasets, demonstrating its potential for biological discovery.
• The system shows remarkable performance in avoiding AI hallucination during data interpretation.
• Provides insightful explanations, system overviews, identification of key regulators, and novel observations.

Conclusions:

• IAN offers a powerful, AI-driven approach to facilitate biological discovery from complex omics data.
• The multi-agent LLM architecture enhances the interpretation of enrichment analysis results.
• The package provides a valuable tool for researchers in bioinformatics and computational biology.