Iterative Gene Enrichment Analysis: interpretable networks for human and AI-assisted biological insights

iGEA transforms over-representation analysis (ORA) outputs into structured, interpretable networks that support human interpretation and AI-assisted exploration. Relative to traditional ORA, which often produces long and fragmented lists of enriched terms, iGEA organizes results as networks, making modules and hub genes easier to identify through visualization and network-based analysis. The paper presents a software framework and demonstrates it on a published HIV dependency factor gene set, representing methodological and case-demonstration evidence.

By iteratively selecting the most significant enriched term and removing its overlapping genes, iGEA yields a compact set of non-overlapping enriched terms within each gene-set collection. This iterative strategy reduces within-collection redundancy, making each collection's output more compact and easier to translate into coherent, testable biological hypotheses. The method is clearly described and produces five compact modules in the HIV dependency factor case, representing a single-case method demonstration.

Integrating collection-specific results yields a cross-collection gene-term network in which genes connect terms from different collections. This network exposes relationships across gene-set collections, which traditional ORA typically struggles to integrate. The paper describes the integration procedure and network structure and shows its output in the HIV dependency factor case, with evidence from method description and case application.

The resulting network structure supports standardized prompting and provides a structured representation for LLM-assisted summarization and exploration of enrichment results. Addressing the lack of reproducibility and statistically supported representations when LLMs summarize enrichment outputs, iGEA provides a structured network as an input basis. The paper presents this as a framework capability and does not report quantitative evaluation of LLM-assisted exploration in the abstract.

The framework targets researchers using ORA to interpret high-throughput gene lists, in settings that require integrating enrichment results across multiple gene-set collections and generating compact modules; its demonstration is based on a published HIV dependency factor dataset, and outputs are intended to support hypothesis generation and interactive exploration rather than to directly establish causal conclusions.

Current evidence comes from abstract-level method description and a single-case demonstration; it remains unclear how iGEA performs across gene lists of different sizes, different gene-set collections, and noisy conditions, and no quantitative evaluation of the reproducibility and statistical support of LLM-assisted summarization is reported; moreover, the abstract does not provide specific statistical details of network construction or benchmark comparisons, so readers assessing robustness should consult the full text figures and supplementary materials.