Inferring transcriptomic programs from circulating tumor DNA methylation signatures in small cell lung cancer using RRBS and EM-seq

Background: Small cell lung cancer (SCLC) exhibits profound epigenetic remodeling and transcriptional plasticity, yet tissue scarcity limits multi-omic profiling. Circulating tumor DNA (ctDNA) methylation analysis via reduced representation bisulfite sequencing (RRBS) or enzymatic methyl-seq (EM-seq) provides a minimally invasive window into tumor biology. However, the relationship between ctDNA methylation features and transcriptomic states such as epithelial-mesenchymal transition (EMT) or neuroendocrine (NE) differentiation remains poorly defined but may serve as means to monitor patient therapeutic response. Methods: We analyzed matched ctDNA methylation RRBS and bulk RNA-seq data from N = 43 SCLC patients collected at baseline. CpG methylation levels were quantified and mapped to gene bodies, promoters, and distal regulatory regions. To capture functional methylation signatures, CpG sites were ranked by their correlation with RNA expression of the corresponding genes and by feature importance predicting expression levels or calculated transcriptomic scores (e.g., EMT, MYC, ASCL1/NEUROD1 subtype indices). High-ranking CpG sites were aggregated into gene-level and pathway-level signatures, followed by GO and KEGG enrichment analyses to identify key regulatory networks. These were further evaluated in SCLC patients treated longitudinally with frontline chemotherapy. Results: Genome-wide ctDNA methylation profiles predicted bulk RNA-seq-derived gene expression as well as predefined score metrics (such as EMT score, NE score) with high concordance (Wilcoxon R² = 0.63 ± 0.07, p < 0.001). Ranking analysis revealed that predictive CpG sites were enriched in enhancer and promoter regions associated with EMT regulators and NE lineage genes. Functionally, GSEA further linked high NE scores to activation of Hedgehog signaling, and EMT scores to enrichment of G2/M checkpoint pathways, suggesting distinct regulatory programs underlying SCLC phenotypes. Conclusions: Integrative analysis of ctDNA methylation and transcriptomic data reveals that ranked, functionally annotated CpG features can accurately infer gene expression programs and biological states in SCLC. These approaches provide a mechanistic framework to interpret ctDNA methylation signatures and enable noninvasive characterization of tumor subtypes and therapeutic resistance in SCLC patients.