A non-invasive circulating tumor DNA methylation classifier to identify benign pulmonary nodules

BACKGROUND: Low-dose computed tomography (LDCT) suffers from a high false-positive rate in the evaluation of pulmonary nodules. Circulating tumor DNA (ctDNA) methylation is a promising complementary biomarker, but its detection is hindered by the highly fragmented nature of ctDNA. METHODS: We developed SAMT-Seq (Single-Strand Amplification Methylation-Targeted Sequencing), optimized for methylation detection in fragmented ctDNA. Lung cancer-specific methylation markers were identified from in-house cohort and TCGA database and validated in paired tissue and plasma samples from 30 early-stage lung cancer patients. A panel of 30 key markers was selected using LASSO regression in a training cohort (n=239). A Gaussian process classifier was developed and validated in two independent cohorts (n=59 and n=207). RESULTS: SAMT-Seq demonstrated superior analytical sensitivity and on-target efficiency compared to a standard commercial Swift method. The 30-marker classifier yielded area under the curve (AUC) of 0.95, 0.95, and 0.92 in the Training Cohort, Validation Cohort 1 and 2, respectively. Notably, it maintained robust performance across nodule types (solid/subsolid), sizes, smoking status, and insitu carcinoma. With a predefined threshold, the model achieved specificity of 100% and 92.16% in Validation Cohort 1 and 2, respectively, suggesting its potential utility in reducing false-positive classifications. CONCLUSIONS: We developed a high-specificity ctDNA methylation classifier that serves as a practical, complementary tool for risk stratification of pulmonary nodules, with the potential to significantly reduce unnecessary invasive procedures. Ongoing prospective diagnostic validation study are evaluating its clinical performance.