Combining Methylated RNF180 and SFRP2 Plasma Biomarkers for Noninvasive Diagnosis of Gastric Cancer
Zhihao Dai,Jin Jiang,Qianping Chen,Minghua Bai,Quanquan Sun,Yanru Feng,Dong Liu,Dong Wang,Tong Zhang,Liang Han,Litheng Ng,Jun Zheng,Hao Zou,Wei Mao,Ji Zhu
DOI: https://doi.org/10.1016/j.tranon.2024.102190
IF: 4.803
2024-01-01
Translational Oncology
Abstract:Introduction Gastric cancer (GC) is a common malignant tumor, and early diagnosis significantly improves patient survival rates. This study aimed to investigate the diagnostic value of ring finger protein 180 (RNF180) and secreted frizzled protein 2 (SFRP2) in GC. Materials & Methods A total of 165 healthy individuals, 34 patients with precancerous gastric lesions, and 104 patients with confirmed GC were divided into training and validation sets; methylated RNF180 and SFRP2 were detected in circulating DNA from blood samples. Six models, including those based on logistic regression, Naive Bayes, K-nearest neighbor algorithm, glmnet, neural network, and random forest (RF) were built and validated. Area under the curve (AUC), sensitivity, specificity, positive predictive value, and negative predictive value were determined. Results In the training set, the RF model with RNF180 and SFRP2 (R + S) had an AUC of 0.839 (95 % CI: 0.727–0.951), sensitivity of 60.3 %, and specificity of 85.5 % for diagnosing GC. The RF model with R + S+ Tumor markers had an AUC of 0.849 (95 % CI: 0.717–0.981), sensitivity of 62.8 %, and specificity of 87.1 %. In the validation set, the RF model with R + S had an AUC of 0.844 (95 % CI: 0.774–0.923), sensitivity of 87.8 %, and specificity of 69.2 %. The RF model with R + S + Tumor markers had an AUC of 0.858 (95 % CI: 0.781–0.939), sensitivity of 85.4 %, and specificity of 76.9 %. Conclusion Our results suggest that RNF180 and SFRP2 could serve as diagnostic biomarkers for GC when using the RF model.