Gastric cancer detection by stomach chromatin–anchored salivary cell-free DNA fragmentomics

<title>Abstract</title> <p>The current liquid biopsy paradigm rests mostly on plasma cell-free DNA (cfDNA), and only to a limited extent on other biofluids. Saliva is molecularly rich but difficult to use, as tumour signal must be separated from abundant background noise. Salivary cfDNA is dominated by shorter fragments, contains substantial microbial DNA, and shows a cancer-associated shift toward longer fragments opposite to plasma. Here we develop a bench workflow and analytical framework to detect gastric cancer from salivary cfDNA. We applied Broad-Range cfDNA Sequencing (BRcfDNA-Seq) to capture fragments lost in conventional library preparation and profiled 281 participants under a PRoBE design, with model fitting restricted to a development cohort (n = 132) and performance evaluated in a separate held-out cohort (n = 149). Applied genome-wide, fragmentomics detected gastric cancer but could not separate signal from background, reaching 95.5% sensitivity at only 37.8% specificity. Anchoring the same coverage to stomach-specific CTCF insulator architecture and refining to discriminative loci raised specificity from 38% to 84.1% (AUROC 0.877, 82% sensitivity); mismatched immune references produced smaller gains, supporting tissue matching. Adding demographic covariates increased AUROC to 0.907 but reduced specificity. Tissue-anchored chromatin analysis converts saliva, a molecularly rich compartment, into a usable substrate for cancer detection.</p>