通过在商业化CGP panel中加入靶向基因调控区域的定制杂交捕获panel，在单次cfDNA检测中结合基因表达与突变谱分析

Introduction: Analysis of cell-free DNA (cfDNA) provides a rapid, repeatable, and non-invasive window into tumor biology, yet tissue biopsies are still often needed for accurate phenotyping and treatment guidance. A liquid biopsy approach that captures both genomic alterations and expression-like features could strengthen patient monitoring and lessen biopsy reliance. Here, we integrate a custom hybrid-capture panel targeting gene regulatory regions with a comprehensive genomic profiling (CGP) panel into a single cfDNA sequencing assay. Our bioinformatic pipeline calculates variant allele frequencies (VAF) and extracts fragmentomic features that reproducibly correlate with gene expression, enabling non-invasive tumor phenotyping.Top of FormBottom of Form Experimental Procedures: Matched FFPE tumor and plasma samples from 31 breast cancer patients were collected and analyzed, along with 11 additional unmatched plasma samples; all patients provided consent. Plasma cfDNA underwent sequencing in a single assay that integrated custom hybrid-capture baits for regulatory regions of 2,704 genes (2.44 Mb) with a CGP panel covering 562 genes (2.4 Mb). Matched FFPE RNA was profiled by RNA sequencing. cfDNA reads were processed through a multi-omic pipeline for quantifying fragmentomic properties across predefined regions assigned to nearby genes and a gene alteration analysis for VAF. Fragmentomic signals were correlated with RNA expression for the same genes. Association of fragmentomic signal, gene expression, and associated tumor biology was assessed. Results: Fragmentomic features with high Pearson correlation to matched tissue RNA-seq were observed and were used to train a breast cancer model comprised of 34 features. Model-selected features showed a more distinctly matched correlation structure to breast tumor tissue RNA-seq than did randomly sampled features. Mapping model-selected genes to 31 breast tumor gene expression ontologies showed an enrichment for genes, such as EZH1 , previously associated with breast cancer. Additional associations with ER status, Ki67%, and HER2 status demonstrated the potential utility of this assay for measuring these markers in blood. Summary and Conclusions: Our findings show that cfDNA fragmentomic patterns captured using both CGP and custom panels, together with AI modeling, can recapitulate gene-level expression traits and enable tissue-free, plasma-based inference of tumor molecular phenotypes. By integrating fragmentomic signals with standard DNA-alteration profiling, hybrid-capture cfDNA sequencing provides two complementary molecular data layers from the same assay. Combining these capabilities can enhance biomarker discovery and may ultimately support clinical decision-making in precision oncology.