TAPS+实现对CSF来源ctDNA和FFPE中CNS肿瘤的直接5mC/5hmC分辨的基因组和甲基化分析

DNA methylation profiling has become a critical tool in the classification of CNS tumors, helping to refine diagnoses when histology or imaging are inconclusive. In parallel, liquid biopsy approaches analyzing circulating tumor DNA (ctDNA) from cerebrospinal fluid (CSF) offer minimally invasive windows into tumor biology, with applications in diagnosis, treatment monitoring, and relapse detection. A unified method that profiles both methylation and somatic alterations from low-input DNA-such as CSF-ctDNA and FFPE-would provide substantial clinical value. TET-Assisted Pyridine-Borane Sequencing (TAPS+) enables such integrated profiling through a bisulfite-free workflow that preserves DNA integrity while detecting CpG methylation, mutations, indels, copy-number changes, and gene fusions from a single library. We applied TAPS+ to 139 CSF samples and 61 FFPE CNS tumors to evaluate analytical performance and clinical utility. Methods: CSF samples were collected through the Perlmutter Cancer Center liquid biopsy program and FFPE CNS tumors from archival tissue. TAPS+ libraries were prepared from 1-100 ng DNA using the Watchmaker Genomics kit and sequenced to 10-80× depth. Somatic variants were called using the nf-gOS pipeline. The Rastair methylation framework was extended to incorporate strand-discordance logic and probabilistic CpG scoring to distinguish true variants from methylation-derived C>T transitions. Results: Across six matched FFPE tumors, Heidelberg classifier confidence improved after applying our CpG-probability and strand-discordance model, increasing from a mean of 0.20 to 0.81, with all samples gaining 0.50-0.79 and low-confidence calls converted to high-confidence predictions. CpG methylation sensitivity improved from 0.78 to 0.99 while maintaining high specificity (0.998→0.991), enabling accurate discrimination of methylation-derived artifacts from true variants. In CSF-derived ctDNA, excluding 11 QC failures, TAPS+ detected ≥1 tissue-confirmed variant in 43 of 67 evaluable samples (64%), while 24 (36%) were ctDNA-negative despite adequate coverage. TAPS+ also recovered canonical CNS tumor drivers across tissue and CSF, including 1p/19q codeletion, trisomy 7, chromosome 10 loss, EGFR and MET amplification in GBM, ERBB2 amplification, IDH pathway mutations, MN1::CXXC5 and EGFRvIII fusions, and MGMT promoter methylation concordant with array-based calls. Conclusions: TAPS+ enables accurate, bisulfite-free profiling of methylation and somatic alterations from low-input and degraded DNA. Improved classifier performance in FFPE tumors and reliable variant recovery in CSF support TAPS+ as a unified genomic-epigenomic assay.