Non-invasive strategy for gastric cancer detection: Integration of cell-free DNA fragmentomics and protein biomarkers.

Abstract

4071 Background: Gastric cancer (GC) ranks the fifth most common cancer worldwide. Although gastroscopy is widely acknowledged as the gold standard for GC detection, its invasiveness limits its utility in large-scale screening. Accurate, non-invasive diagnostic modalities are urgently needed. Cell-free DNA (cfDNA) fragmentomics has emerged as a promising tool for cancer detection. This study aimed to develop and validate a novel non-invasive model for GC detection based on cfDNA fragmentomics and protein biomarkers. Methods: A total of 257 participants, comprising 120 GC patients and 137 non-cancer individuals, were enrolled and divided into a training cohort (n = 129) and a test cohort (n = 128). Low-coverage whole-genome sequencing was performed on plasma cfDNA. We evaluated eight multi-dimensional cfDNA fragmentomics features individually, and selected four, including fragment size ratio, copy number variation, 9-bp end motif, and fragment size at transcription starting sites, to build an ensemble model (GaFraD) to detect GC. Furthermore, conventional protein biomarkers of GC were assessed and integrated with GaFraD model to generate a CONFIRM model aimed at improving diagnostic performance. Results: The GaFraD model achieved an area under the receiver-operating characteristic curve (AUC) of 0.953 (95% CI: 0.918 – 0.979) in the training cohort and 0.970 (95% CI: 0.944 – 0.990) in the test cohort, with a sensitivity of 95.0% and a specificity of 80.9%. By incorporating protein biomarkers CA19-9 and PG-I/PG-II, the CONFIRM model further reached an AUC of 0.967 (95% CI: 0.929 – 0.992) in the training cohort and 0.986 (95% CI: 0.966 – 1.000) in the test cohort, attaining a sensitivity of 95.0% and specificity of 95.6%. Moreover, the CONFIRM model also exhibited remarkable performance in discriminating early-stage GC patients from controls (AUC = 0.983, sensitivity: 95.6%, specificity: 94.2%). Conclusions: Our model exhibits high discriminatory power in distinguishing GC patients from controls, highlighting the strong clinical potential of combining cfDNA fragmentomics with protein biomarkers for non-invasive, early GC detection. This approach offers a promising pathway towards earlier, accurate, and non-invasive clinical diagnosis of GC.