Incomplete Concordance Between Nominal Eosinophilic Labels and Molecular Burden in Chronic Rhinosinusitis with Nasal Polyps

Background/Objectives: Chronic rhinosinusitis with nasal polyps (CRSwNP) is a heterogeneous inflammatory disease in which eosinophilic subclassification is widely used for clinical stratification. However, it remains unclear how closely nominal histologic eosinophilic labels reflect the broader molecular organization of diseased tissue. Methods: We performed an inference-based integrative analysis of public datasets spanning discovery single-cell RNA sequencing (scRNA-seq), independent scRNA-seq validation, GeoMx digital spatial profiling, and bulk transcriptomic replication cohorts. A sample-level molecular burden framework was constructed using four dimensions: type 2 inflammation, epithelial injury/remodeling, extracellular-matrix remodeling, and barrier/defense impairment. Composite burden and component-level features were then examined across nominal eosinophilic categories, epithelial states, spatial compartments, and independent bulk cohorts. Results: Nominal eosinophilic labels were directionally informative but incompletely concordant with molecular burden. In the discovery cohort, eosinophilic CRSwNP samples were enriched toward the higher-burden end, whereas nominally non-eosinophilic CRSwNP samples extended across the intermediate-to-high burden range. Across discovery and validation scRNA-seq datasets, GeoMx spatial profiling, and independent bulk cohorts, the most reproducible burden-associated signals centered on epithelial injury/remodeling-like programs and related remodeling features. In the epithelial compartment, higher burden was associated with epithelial state reorganization, stronger wounding-associated activity, and trajectory-linked glandular/secretory remodeling. Independent validation and spatial analyses further supported epithelial wounding-, barrier-, and myeloid remodeling-related features, whereas type 2 context signals were directionally consistent but less uniform across platforms. In bulk replication, composite burden, epithelial wounding, and myeloid remodeling were more consistent across cohorts than type 2 context alone. Conclusions: Nominal eosinophilic labels in CRSwNP capture clinically relevant but incomplete information about underlying tissue biology. Epithelial injury/remodeling-like programs and remodeling-linked myeloid features emerged as the most stable organizational axes of molecular burden across public multimodal datasets. These findings support a graded, multidimensional view of CRSwNP and may complement, rather than replace, conventional pathology-based eosinophilic subclassification.