How nih-nei/ian works
Project type and purpose
• This is an R-package (named “IAN”) that implements a modular pipeline for integrated “omics” data analysis.
• It’s designed to take a list of differentially expressed genes (DEGs) as input, perform multiple enrichment and interaction analyses, leverage a multi-agent LLM system for interpretation, and produce interactive network visualizations and an HTML report.
File structure
Top-level files:
• DESCRIPTION, NAMESPACE: define the R package metadata and exports
• R/: contains the core R code, one .R file per logical module
• man/: auto-generated Rd documentation for every function and class
• inst/docs/: supplementary scripts (Python, grounding, evaluation), parameter guides, example data
• inst/rmd/: report template and styling (R Markdown)
Key patterns:
• Each R/IAN_*.R file represents a distinct processing stage (gene mapping, enrichment, prompt generation, network generation, visualization)
• There is a central orchestration module (IAN_multi_agent_system.R) that ties together sub-modules
• Deployment artifacts include rmarkdown templates and CSS for report generation
Extract architectural insights from the README
• Inputs: custom DEG lists or outputs from DESeq2/Seurat
• Enrichment backends: clusterProfiler (GO, Reactome, KEGG), enrichR (ChEA), STRINGdb for PPI
• Multi-agent LLM system: Google Gemini API with prompt generators and revision loops
• Outputs: visNetwork interactive graphs, HTML reports via rmarkdown