Abstract
BACKGROUND: Molecular monitoring of Philadelphia-negative myeloproliferative neoplasms (MPN) relies on genomic DNA (gDNA) from peripheral blood (PB). As variant allele frequency (VAF) reduction correlates with clinical outcomes, optimizing monitoring strategies has become important. However, whether gDNA accurately reflects bone marrow (BM) clonal dynamics and whether cell-free DNA (cfDNA) provides complementary information remains undefined. METHODS: We compared cfDNA and PB gDNA using methylation-based deconvolution, a cell line-derived xenograft (CDX) model, and paired clinical samples from two independent MPN cohorts (technical, n=7 with matched BM; validation, n=40 with longitudinal monitoring). RESULTS: Methylation deconvolution showed an enrichment of BM-derived DNA in cfDNA (47.4% vs 18.9%, P = 0.035), providing exploratory evidence of its origin. The CDX model indicated that cfDNA serves as a superior surrogate for BM clonal burden, whereas gDNA underestimated it (P < 0.001). Clinically, mutation VAF showed high concordance (r=0.887, P < 0.001), yet cfDNA detected higher VAF (+22.6% mean advantage). Gene-specific analysis revealed CALR mutations had the 59.3% cfDNA advantage (P = 0.009). During treatment monitoring, cfDNA showed a trend toward higher sensitivity for resistance detection (25.6% vs. 17.9%), whereas gDNA appeared to be more sensitive for response confirmation (23.1% vs. 15.4%). Lymphocyte percentage emerged as a novel predictor of cfDNA advantage (r = 0.765, P = 0.045 in technical; r = 0.509, P < 0.001 in validation). CONCLUSIONS: While constrained by sample size and specific treatment contexts, our observations suggest that cfDNA and gDNA provide complementary value for MPN monitoring. Supported by preclinical models and technical analysis, cfDNA more effectively reflects BM clonal burden, while longitudinal observations suggest its capacity to track treatment-related clonal dynamics. In contrast, gDNA is more informative in confirming treatment response. The lymphocyte percentage predicts cfDNA utility, enabling rational test selection and a practical framework for optimizing MPN management.