A prediction model based on DNA methylation biomarkers and radiological characteristics for identifying malignant from benign pulmonary nodules

Abstract Background Lung cancer remains the leading cause of deaths across world. Early detection lung by low-dose computed tomography (LDCT) can reduce mortality rate. However, making a definitive preoperative diagnosis malignant pulmonary nodules (PNs) found LDCT is clinical challenge. This study aimed to develop prediction model based on DNA methylation biomarkers and radiological characteristics for identifying from benign PNs. Methods We assessed three ( PTGER4 , RASSF1A, SHOX2 ) clinically-relevant variables in training cohort 110 individuals with Four machine-learning-based models were established compared, including K-nearest neighbors (KNN), random forest (RF), support vector machine (SVM), logistic regression (LR) algorithms. Variables best-performing algorithm selected through stepwise use Akaike’s information criterion (AIC). The constructed was compared Mayo Clinic using non-parametric approach DeLong et al. area under receiver operator characteristic curve (AUC) analysis. Results A finally one developed achieved an AUC value 0.951 PNs diagnosis, significantly higher than (0.912, 95% CI:0.843–0.958, p = 0.013) or (0.823, CI:0.739–0.890, 0.001). Validation testing 100 subjects confirmed diagnostic value. Conclusion have shown that integrating could more accurately identify CT-found our may provide utility combination improve over-all cancer.