Integrative multi-omics and radiogenomic profiling decodes NNK-related tumor remodeling and prognostic stratification in pancreatic cancer

Objective: To elucidate how the tobacco-specific carcinogen 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone (NNK) relates to pancreatic cancer (PC) progression and to develop a CT-based radiomics model that non-invasively decodes NNK-related molecular alterations. Methods: NNK-related genes were identified by integrating toxicogenomic mining across multiple databases with multi-cohort transcriptomic analyses. Differential expression analysis, weighted gene coexpression network analysis (WGCNA), a multi-algorithm machine-learning framework, and Cox proportional hazards modeling were applied to screen key regulators and prognosis-related factors. Single-cell RNA sequencing was performed to characterize cellular heterogeneity and tumor–stroma communication associated with ITGA3. Radiomic features were extracted from CT scans, and a radiomics signature was constructed. Prognostic value was evaluated in the cancer imaging archive (TCIA), and an independent surgical cohort. Results: A total of 268 NNK-associated PC genes were identified, enriched in extracellular matrix (ECM) remodeling, focal adhesion, PI3K-AKT signaling, and DNA-damage response pathways. Machine-learning prioritization yielded 15 key candidates, among which ITGA3 remained an independent prognostic factor ( P < 0.05). Single-cell analysis showed enrichment of basal/epithelial–mesenchymal transition (EMT)-high epithelial subsets and intensified epithelial–cancer-associated fibroblast–macrophage communication in ITGA3-high tumors. The NNK-related ITGA3-guided radiomics model predicted ITGA3 expression (AUC = 0.893 and 0.839) and achieved significant postoperative survival stratification. Conclusion: NNK is associated with a malignant transcriptional program characterized by epithelial remodeling, stromal interaction, and immune suppression. These molecular features can be captured non-invasively through CT-based radiomics. The NNK-related ITGA3-associated radiomics (NIR) score provides a practical tool for preoperative risk stratification in PC and offers a bridge linking environmental carcinogen exposure with tumor imaging phenotypes.