Integrated transcriptomic and clinical prediction model for malignant pulmonary nodules: Model development, external dataset evaluation, and single-center expression validation

Discriminating between benign and malignant pulmonary nodules remains a clinical challenge, complicating treatment decisions. This study developed a predictive model integrating transcriptomic and clinical data to characterize pulmonary nodules. Gene expression profiles and clinical data from GSE108375 were analyzed. Differential expression and weighted gene co-expression network analysis (WGCNA) were performed. Patients were randomly split into training and validation sets (7:3). The model was further evaluated using an external dataset (GSE225787), and candidate gene expression was examined in a single-center clinical cohort (n = 30) using blood reverse‑transcription quantitative real‑time PCR and nodule-tissue immunohistochemistry (IHC). WGCNA identified the MEgreen module as significantly associated with malignant nodules. Among 776 differentially expressed genes between malignant and benign nodules, 36 overlapped with the MEgreen module. Last absolute shrinkage and selection operator selected age, lesion size, and 5 genes (RBX1, GZMK, AKR1C3, NDUFA4, CELA1) for model construction. The model achieved AUCs of 85.16% in the training set and 80.23% in the internal validation set, with an AUC of 0.867 in the external dataset. In the single-center clinical cohort, blood CELA1 and NDUFA4 mRNA levels were significantly higher in the malignant group, and tissue IHC measurements of the corresponding proteins also differed significantly between groups. The integrated transcriptomic and clinical model showed potential for estimating malignancy risk in pulmonary nodules. The expression findings for CELA1 and NDUFA4 support further investigation of these candidate markers in larger clinical cohorts.