BACKGROUND

DNA methylation was suggested as the promising biomarker for lung cancer diagnosis. However, it is a great challenge to search for the optimal combination of methylation biomarkers to obtain maximum diagnostic performance.

RESULTS

In this study, we developed a panel of DNA methylation biomarkers and validated their diagnostic efficiency for non-small cell lung cancer (NSCLC) in a large Chinese Han NSCLC retrospective cohort. Three high-throughput DNA methylation microarray datasets (458 samples) were collected in the discovery stage. After normalization, batch effect elimination and integration, significantly differentially methylated genes and the best combination of the biomarkers were determined by the leave-one-out SVM (support vector machine) feature selection procedure. Then, candidate promoters were examined by the methylation status determined single nucleotide primer extension technique (MSD-SNuPET) in an independent set of 150 pairwise NSCLC/normal tissues. Four statistical models with fivefold cross-validation were used to evaluate the performance of the discriminatory algorithms. The sensitivity, specificity and accuracy were 86.3%, 95.7% and 91%, respectively, in Bayes tree model. The logistic regression model incorporated five gene methylation signatures at AGTR1, GALR1, SLC5A8, ZMYND10 and NTSR1, adjusted for age, sex and smoking, showed robust performances in which the sensitivity, specificity, accuracy, and area under the curve (AUC) were 78%, 97%, 87%, and 0.91, respectively.

CONCLUSIONS

In summary, a high-throughput DNA methylation microarray dataset followed by batch effect elimination can be a good strategy to discover optimal DNA methylation diagnostic panels. Methylation profiles of AGTR1, GALR1, SLC5A8, ZMYND10 and NTSR1, could be an effective methylation-based assay for NSCLC diagnosis.