A blood-based transcriptomic signature for noninvasive diagnosis of gastric cancer

Abstract

Background Delayed detection of tumours contributes to poor prognosis in patients with gastric cancer (GC). The invasive nature of endoscopy and the absence of an effective serum markers highlight the need to develop novel, noninvasive biomarkers. Methods We performed biomarker discovery and validation to identify candidate genes in three gene expression data sets. After validating the gene panel in clinical tissues, we translated the gene panel into serum samples by performing training and validation in 89 samples from GC patients and 54 from healthy donors in two independent cohorts. Results We identified a nine-gene panel in the discovery phase, with subsequent validation in tissue specimens. Using a serum training cohort, we developed a 5-gene risk prediction formulae for the diagnosis of GC; bootstrapped analysis exhibited an AUC of 0.896. We validated this 5-gene biomarker panel using an independent serum cohort, yielding an AUC of 0.947. This biomarker panel successfully identified GC, regardless of tumour histology. Notably, biomarker performance for detection of stage 1 and 2 GC displayed an AUC of 0.928 and 0.980 in both serum cohorts. Conclusions We identified a novel 5-gene biomarker panel for noninvasive diagnosis of GC, which might serve as a potential diagnostic tool for early detection.