Oesophageal adenocarcinoma (OAC) is among the most aggressive malignancies, with a five-year survival rate of less than 20%. We aimed to identify prognostic protein biomarkers for overall survival (OS) and molecular-based characterisation of OAC using mass spectrometry (MS)-based proteomics. Two independent cohorts were analysed, encompassing 241 OAC patients, over 600 samples, and 1,319 MS runs. Based on differential abundance analysis (malignant vs normal adjacent) followed by Cox regression analyses with LASSO regularisation, a five-protein prognostic signature for OS was identified (HR 20.6 [6.8, 61.6], p-value <0.001). The signature was validated in an independent cohort treated with neo-adjuvant treatments (HR 1.3 [1.0, 1.6], p-value=0.02). The signature was independent of other clinical variables, including tumour size, lymph nodes status, grade, and age.
Proteomics-based unsupervised consensus clustering identified four novel OAC subtypes with distinct protein expression and molecular mechanisms, with Subtype-I displaying keratin-related functions, Subtype-II presenting collagen expression, Subtype-III showing neutrophil degranulation, and Subtype-IV indicating the activation of amino acid metabolism and translation. Incorporation of the proteomic subtypes into the five-protein signature in a multivariate Cox model significantly enhanced prediction, achieving a receiver operator curve area under the curve (ROC AUC) of 0.78. In conclusion, we identified four proteomic subtypes and a set of five proteins that significantly identify high-risk patients for OS; these findings would enable better-informed treatment decisions and thus improve patients’ outcomes.