Updated project metadata. Timely diagnosis of gastric adenocarcinoma (GAC) can effectively prevent the deterioration of the disease and significantly improve the survival rate of patients. Currently, tumor markers used in clinical practice cannot meet the needs of patients, and it is urgent to develop biomarkers with high sensitivity and specificity to guide clinical diagnosis. In this study, serum samples from 33 patients with GAC and 19 healthy persons were collected. EVs were extracted by ultracentrifugation and EV protein expression profiles were obtained by DIA quantitative mass spectrometry. Multi-group differential protein expression analysis showed that 23 intersecting proteins had the same expression trend, among which 15 EV proteins were chosen as candidate biomarkers for GAC diagnosis. Afterward, a subset of 2 to 6 proteins was randomly selected as features for logistic regression modeling. To verify the diagnostic performance of these models, serums from a new cohort of 12 patients with gastric cancer and 18 healthy controls were collected, and the EV protein were quantified using the same method. We finally found a panel consisting of six proteins that performed the best as classifier. We also tried to discover biomarkers for diagnosis of advanced stage in GAC, and the results showed a three-protein panel had the best classification performance. In conclusion, we identified new protein biomarker panels from serum EVs for early diagnosis of gastric cancer that worth further validation.