Deep functional interpretation of influenza-induced ciliary gene dysregulation through stepwise Large Language Model profiling 2302296

Abstract Introduction The interpretation of large-scale transcriptional data remains a significant challenge in functional genomics, particularly in complex biological contexts such as host-pathogen interactions. Methods We present a systematic approach combining air-liquid interface (ALI) cultures with stepwise Large Language Model (LLM) analysis to achieve deep functional interpretation of ciliary gene regulation during influenza infection. From 2,828 differentially expressed genes, initial high-throughput LLM screening identified 29 genes with high confidence scores specifically associated with ciliated cell biology. We conducted detailed functional profiling of these candidates through human-in-the-loop validation. Results Analysis revealed a coordinated program of ciliary gene dysregulation. Key genes, including DNAH5, DYNC2H1, and DNAAF4-CCPG1, showed consistent downregulation by 24-48 hours post-infection. This pattern was validated across independent datasets from Influenza, Rhinovirus and SARS-CoV-2 infections, suggesting a conserved mechanism of mucociliary clearance impairment across respiratory viral infections. Conclusion By focusing our analysis on ciliated cell-associated genes, we uncovered specific mechanisms of viral pathogenesis while establishing a generalizable framework for context-aware interpretation of complex biological datasets. Funding Source National Institute of Allergy and Infectious Diseases (NIAID) Topic Categories Computational and Systems Immunology (COMP)