Fig. 4

Tissue-derived DMMs had a limited contribution in detecting early-stage HCC in cfDNA (training cohort). A The ROC curve revealed AUC for 11 tissue-derived DMMs in distinguishing cirrhotic HCC from cirrhosis (tissue). The top 5 DMMs selected for cfDNA sample testing were TSPYL5, BOP1, SPAG6, NRIP3, and FOXD3. B qMSP-based methylation levels relative to the ACTB reference gene were shown for the 5 selected DMMs in 91 cfDNA samples (29 late-stage HCC, 30 early-stage HCC, and 32 cirrhosis). All 5 DMMs exhibited hypermethylation in late-stage HCC compared to cirrhosis, while only NRIP3 showed statistical significance between early-stage HCC and cirrhosis (Mann–Whitney U test; ns, not significant; **, p < 0.01; *, p < 0.05). C Performance of DMMs and ASAP/GAAD in HCC detection. Binary logistical regression (Enter method) was performed to create predictive formulas for the combined 5 DMMs comprising TSPYL5, BOP1, SPAG6, NRIP3, and FOXD3, as well as for the combination of ASAP/GAAD and 5 DMMs. Additionally, binary logistical regression (backward elimination) was utilized to develop an alternative algorithm for the combination of ASAP/GAAD and 5 DMMs, resulting in ASAP/GAAD and FOXD3. The p-value for the ROC curve was compared between the combined model and the ASAP/GAAD score