The value of an integrated multi-omics model in the diagnosis of benign and malignant pulmonary nodules

Abstract

Background: In recent years, multi-omics models based on a variety of biomarkers have been continuously developed and increasingly applied in the field of oncology, especially in the early diagnosis of lung cancer. This study aimed to integrate computed tomography (CT) radiomics with seven lung cancer-associated autoantibodies (AABs) to develop multi-omics predictive models for pulmonary nodule (PN) characterization. Methods: This retrospective study enrolled 179 patients with PNs measuring from 5 to 30 mm in diameter who underwent thoracic surgery at Zhongda Hospital, Southeast University between January 2020 and December 2024. The patients were pathologically categorized into lung cancer (n=87) and non-lung cancer (n=92) groups, and then randomly allocated into training and test sets at a ratio of 7 to 3. Least absolute shrinkage and selection operator (LASSO) regression was used for feature screening to construct a clinical model based on five clinical characteristics. A radiomics prediction model was constructed based on the radiomics features identified after delineating the regions of interest and extracting the radiomics features; the rad-score for each patient was calculated to develop a multi-analytic comprehensive model by combining different markers. The diagnostic performances of the models were compared using the area under the curve (AUC), accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value. Results: The multi-omics model demonstrated superior diagnostic accuracy with an AUC of 0.902 [95% confidence interval (CI): 0.817–0.986], accuracy of 82.4%, sensitivity of 88.5%, and specificity of 80.0%, outperforming the clinical (AUC =0.848; 95% CI: 0.777–0.919) and radiomics (AUC =0.854; 95% CI: 0.786–0.922) models. Notably, the radiomics model exhibited high sensitivity (96.6%) but poor specificity (63.6%), while the multi-omics model resolved this trade-off via the synergistic integration of clinical-radiomic-biomarker features, achieving significant improvements in the PPV (81.5% vs. 72.7%) compared to the clinical model. Conclusions: Integrating CT radiomics with seven lung cancer-AABs established a robust multi-omics framework for PN diagnosis. Compared to the standalone clinical or radiomics models, this comprehensive model demonstrated superior diagnostic performance.