Regression-based clustered CpG methylome profiling reveals divergent epigenetic trajectories in paired primary and recurrent hepatocellular carcinoma

Recurrent hepatocellular carcinoma (HCC) can arise with variable molecular similarity to the primary tumor, but practical approaches for quantifying methylome divergence in paired tumors remain limited. We performed reduced representation bisulfite sequencing on paired primary and recurrent HCC lesions and matched normal margins from ten patients, quantifying clustered CpG methylation across 77,790 MspI-defined loci. Regression-based methylation shift patterns (MSPs) captured tumor-specific curve shapes, and Euclidean distances between regression-coefficient vectors grouped tumors into four MSP classes. We further introduced tH, a triangle-area metric summarizing within-patient methylome divergence between primary and recurrent tumor-associated MSPs. In this exploratory cohort, tH distinguished lower- and higher-divergence recurrence patterns that showed concordance with serum marker patterns, mutation-based tumor relatedness, and recurrence-free survival differences. Interval-based analyses revealed prominent hypermethylation at low-methylation loci, hypomethylation at high-methylation loci, and heterogeneous changes in intermediate methylation tiers. These findings establish an MSP/tH framework for quantifying methylome divergence between paired primary and recurrent HCCs and for stratifying tumors by shared epigenetic shift patterns, providing a basis for future validation in larger paired-tumor cohorts.