| Description |
Gene expression patterns of tumor cells can be inferred from features of circulating cell-free DNA (cfDNA), such as histone modifications and promoter fragmentation. However, the direct link between patterns in the cfDNA and tumor-specific chromatin accessibility has not been experimentally established and its application is limited to the diagnosis of cancer. Therefore, we surveyed plasma cfDNA from 500 individuals including patients with 21 disease types and healthy donors. We found that cfDNA TSS coverage accurately distinguished patients with stomach adenocarcinoma (STAD, AUC 0.96), breast cancer (BRCA, AUC 0.87), and pancreatic ductal cancer (PDAC, AUC 0.88) from healthy donors, patients with gastritis, noncancer illness, and other cancer types. Cancer-specific cfDNA open chromatin features were subsequently validated by bulk ATAC-seq, single-cell ATAC, and RNA-seq analyses. Using a capture assay targeting the cancer-specific TSSs followed by high-throughput sequencing, the AUC reached 0.88 for an independent cohort of 80 plasma samples. Deconvolution of cfDNA signals indicated the appearance of tumor cells and a decline in the proportion of dendritic cells (DCs) in the blood of patients with PDAC compared to that of healthy donors. Moreover, the cfDNA-deconvoluted cell-type fractions were highly correlated with the cell fractions of the Peripheral Blood Mononuclear Cell (PBMC) identified with scRNA-seq (Pearson correlation 0.97). Together, our results revealed the potential of using cfDNA-inferred chromatin accessibility not only to accurately identify patients with cancer but also to identify disease-specific and pathology-associated transcriptional programs from blood samples. |
