Establishment of Two Data Mining Models of Lung Cancer Screening Based on Three Gene Promoter Methylations Combined with Telomere Damage

Abstract

Objective To identify the significance of a support vector machine (SVM) model and a decision tree (DT) model for the diagnosis of lung cancer combined with the detection of fragile histidine triad (FHIT), RAS association domain family 1 (RASSF1A) and cyclin-dependent kinase inhibitor 2A (p16) promoter methylation levels and relative telomere length (RTL) of white blood cells from peripheral blood DNA. Methods The levels of p16, RASSF1A and FHIT promoter methylation and the RTL of white blood cells in peripheral blood DNA of 200 healthy individuals and 200 lung cancer patients were analyzed by SYBR Green-based quantitative methylation-specific PCR and quantitative PCR. Based on the 4 biomarkers, SVM and DT models were developed. Results The levels of FHIT, RASSF1A and p16 promoter methylation were 3.33 (1.86-6.40) and 2.85 (1.39-5.44) (p = 0.002); 27.62 (9.09-52.86) and 17.17 (3.86-50.87) (p = 0.038); and 0.59 (0.16-4.50) and 0.36 (0.06-4.00) (p = 0.008) in cases and controls, respectively. RTL was 0.93 ± 0.32 and 1.16 ± 0.57 (p<0.001). The areas under the receiver operating characteristic (ROC) curves of the Fisher discriminant analysis, SVM and DT models were 0.670 (0.569-0.761), 0.810 (0.719-0.882) and 0.810 (0.719-0.882), respectively. Conclusions The SVM and DT models for diagnosing lung cancer were successfully developed through the combined detection of p16, RASSF1A and FHIT promoter methylation and RTL, which provided useful tools for screening lung cancer.