Background In solid tumor Phase 1/2 trials (NCT02407990; NCT04068519), tislelizumab demonstrated clinical benefit, including in advanced gastroesophageal adenocarcinoma (GEA). However, the majority of patients with GEA did not respond, highlighting the need to understand mechanisms of resistance and identify predictive biomarkers for response. Methods All tislelizumab-treated patients with GEA from the Phase 1/2 trials were included ( N  = 105). Programmed death-ligand 1 (PD-L1) expression (Tumor Area Positivity TAP ≥ 5%), interferon gamma (IFN γ )-related gene signature, gene expression profile, tumor mutational burden (TMB), and gene hyperamplification (HA) were analyzed for correlation with tislelizumab. Results A moderate association was observed between PD-L1 TAP ≥ 5%, IFN γ gene signature, TMB-high and efficacy. A potential correlation between hyperamplification (HA +) and worse outcomes with programmed cell death protein 1 (PD-1) inhibition was identified. Hyperamplified genes were mainly enriched in cancer progression pathways, including cell cycle and RTK-RAS-PI3K pathways. Joint PD-L1 TAP ≥ 5% and lack of hyperamplification showed the most favorable benefit with an objective response rate of 29.4%, and median progression-free survival and overall survival of 4.1 and 14.7 months, respectively. Tumors with TAP ≥ 5% and HA − had inflamed immune signatures with increased immune cell infiltration, enhanced anti-tumor cytotoxic activity and antigen presentation signatures. Findings were validated in two independent gastric and gastrointestinal cancer cohorts treated with immune checkpoint inhibitors. Conclusions In GEA, PD-L1 positivity, IFN γ -related gene signature and TMB-high status were positively associated with tislelizumab clinical benefit, whereas HA was associated with worse clinical outcomes. Combining PD-L1 positivity and HA − may help identify patients more likely to benefit from PD-1 blockade.