Background: Gastric cancer (GC) is the third leading cause of cancer death. Early detection is a key factor to reduce its mortality. Methods: We retrospectively collected pre- and postoperative serum samples as well as tumour tissues and adjacent normal tissues from 100 GC patients. Serum samples from non-cancerous patients were served as controls (n = 50). A high-throughput protein detection technology, multiplex proximity extension assays (PEA), was applied to measure levels of over 300 proteins. Alteration of each protein was analysed by univariate analysis. Elastic-net logistic regression was performed to select serum proteins into the diagnostic model. Findings: We identified 19 serum proteins (CEACAM5, CA9, MSLN, CCL20, SCF, TGF-alpha, MMP-1, MMP-10, IGF-1, CDCPI, PPIA, DDAH-1, HMOX-1, FLI1, IL-7, ZBTB-17, APBB1IP, KAZALD-1, and ADAMTS-15) that together distinguish GC cases from controls with a diagnostic sensitivity of 93%, specificity of 100%, and area under receiver operating characteristic curve (AUC) of 0.99 (95% CI: 0.98-1). Moreover, the 19-serum protein signature pro-vided an increased diagnostic capacity in patients at TNMI-II stage (sensitivity 89%, specificity 100%, AUC 0.99) and in patients with high miaosatellite instability (MSI) (91%. 98%, and 0.99) compared to individual proteins. These promising results will inspire a large-scale independent cohort study to be pursued for validating the proposed protein signature. Interpretation: Based on targeted proteomics and elastic-net logistic regression, we identified a 19-serum protein signature which could contribute to clinical GC diagnosis, especially for patients at early stage and those with high MSI. (C) 2019 The Authors. Published by Elsevier B.V.