Gianetto, Q. G.; Michel, V.; Douche, T.; Nozeret, K.; Zaanan, A.; Colussi, O.; Trouilloud, I.; Pernot, S.; Ungeheuer, M.-N.; Julie, C.; Jolly, N.; Taieb, J.; Lamarque, D.; Matondo, M.; Touati, E.

Background: Gastric cancer (GC) is often associated with a poor prognosis, due to its asymptomatic early stages. The current gold standard for diagnosing GC, upper endoscopy, is invasive and has limited sensitivity for detecting gastric preneoplasia such as dysplasia. Non-invasive biomarkers, such as circulating proteins in the blood, hold promise for the early detection of asymptomatic gastric lesions. Methods: During an exploratory study, plasma samples from 39 participants, including patients diagnosed with different gastric pathologies (gastritis, preneoplasia, GC) and healthy controls, were analyzed using mass spectrometry-based proteomics. Data are available via ProteomeXchange with identifier PXD062315. Differentially abundant proteins were identified through pairwise comparisons and sparse Partial Least Squares Discriminant Analysis (sparse PLS-DA). Fifteen candidate proteins were selected and quantified by ELISA in plasma samples from the full cohort of 138 participants. Results: Proteomic profiling identified 691 proteins in plasma samples. Pairwise comparisons highlighted up to 213 candidate biomarkers capable of distinguishing cancer patients from healthy controls, while distinguishing gastritis and preneoplasia proved challenging due to their similar proteomic profiles. Sparse PLS-DA identified 85 proteins that distinguish all patient groups. Subsequent ELISA validation confirmed Leptin as a promising biomarker for detecting gastric preneoplasia, particularly in women. Additionally, nine additional proteins (ATAD3B, IGFALS, JUP, LBP, MAN2A1, ARG1, CA2, HPT, KRT14) showed differential plasma levels across patients groups, with variations influenced by age and gender. Multi-biomarker prediction models incorporating factors such as age, gender, and H. pylori status, outperformed single-protein models. Using 4-fold cross-validation repeated 10 times, the top-performing models achieved high predictive accuracy, with mean AUROC values of 85.3% for classifications cancer vs. non-cancer and 83.9% for cancer/preneoplasia vs. healthy/gastritis. On the full cohort, optimal biomarker combinations reached AUROC values exceeding 94% for these classifications. Conclusions: This study introduces a novel, non-invasive approach for predicting GC based on plasma protein concentrations across a broad spectrum of gastric pathologies. Predictive models, validated through robust cross-validation, demonstrated high accuracy using a limited panel of biomarkers. Leveraging simple blood sampling, this strategy holds promise for high-risk gastric mucosal lesions, including at asymptomatic stages. Such an approach could significantly improve early detection and clinical management of GC, offering direct benefit for patient outcomes.