Abstract A22: Establishing a deep cfDNA methylation sequencing-based signature for noninvasive early-stage lung cancer diagnosis

Abstract

Abstract Although mortality rate has been greatly reduced by low-dose computed tomography (LDCT) screening of high-risk patients, the benefit is compromised by the high false-positive rate, thus necessitating the development of biomarkers differentiating benign and malignant nodules. The role of aberrant DNA methylation in the process of tumorigenesis both at individual genes and a genome-wide scale has been well elucidated. It occurs very early in cancer development, and thus is capable of serving as a diagnostic biomarker. In this prospective study, we evaluated the feasibility and performance of assessing cfDNA methylation status as an early diagnostic marker for lung cancer. We developed a novel targeted bisulfite sequencing assay called ELSA and designed a methylation panel based on TCGA data on 838 lung cancer patients and 74 normal controls. We then interrogated the methylation status of 7,572 regions consisting of about 100,000 CpG sites from plasma samples of 163 newly recruited Chinese patients with early-stage NSCLC and 41 with benign nodules using our panel. A region-based methylation statistic was used in the classification model. We constructed diagnostic classification models using elastic net and random forest based on 3,000 DMRs selected by logistic regression between benign and malignant nodules using FFPE samples. Within this cohort, 54 patients had malignant tumor (39 stage I, 6 stage II and 9 stage IIIa) and 30 patients had benign nodules. We performed 5-fold cross-validation with 20 time repeats to gain a robust estimation of model performance, achieving a sensitivity of 92.6%, specificity of 100% and AUC of 98.1%. We further refined the DMRs selected above using 68 healthy baseline plasmas, and subsequently constructed a weighting decision tree model for plasma samples obtained from a separate cohort consisting of 163 patients with malignant tumor (127 stage I, 36 stage II and above) and 41 patients with hamartoma, chronic inflammation, granuloma, or other lung benign lesions, achieving a sensitivity of 84.0%, specificity of 85.7% and AUC of 90.7% based on 170 markers. We further analyzed PPVs according to stage and tumor size. PPVs for patients with stage I, II, and III diseases are 82.4%, 84.6%, and 94.4%, respectively; and for tumor diameter less than 1 cm, between 1-2 cm, and greater than 2 cm are 78.5%, 87.1%, and 88.2%, respectively. Citation Format: Zhihong Zhang, Naixin Liang, Wanglong Deng, Chenyang Wang, Tao Zheng, Wenjun Li, Jian Hu, Jinh Su, Han Han-Zhang, Bingsi Li, Hao Liu, Shanqing Li. Establishing a deep cfDNA methylation sequencing-based signature for noninvasive early-stage lung cancer diagnosis [abstract]. In: Proceedings of the Fifth AACR-IASLC International Joint Conference: Lung Cancer Translational Science from the Bench to the Clinic; Jan 8-11, 2018; San Diego, CA. Philadelphia (PA): AACR; Clin Cancer Res 2018;24(17_Suppl):Abstract nr A22.