52 Background: Molecular heterogeneity of gastric cancer discerns patient survival in ways that clinically-based prediction models cannot. To date, individual microRNAs(miRs) expressed by gastric tumors have been associated with survival. We offer a unique molecular-based method of prognosticating the survival of gastric cancer patients with an efficient panel of miRs. Methods: From The Cancer Genome Atlas(TCGA), we studied subjects with gastric adenocarcinoma, who had undergone R0-R1 resection and had data on clinical characteristics, overall survival (OS), and miR expression. Of miRs quantified by TCGA, miRs expressed by at least 15% of subjects’ tumors were evaluated as continuous variables. From 10 replicate samples, each containing 80% of current subjects, miRs were selected using proportional hazards regression adjusted for age with stepwise selection. Cross-validated miRs were retained for the panel if they optimized an accelerated failure-time model of OS using the full cohort. Results: Among subjects (n=254, age 64.7+10.9 years), median OS was 2.38 (95% Confidence Interval: 2.09-3.85) years. Of the 913 evaluable miRs, 10 distinct miRs (miR-373, miR-548ay, miR-659, miR-891a, miR-1243, miR- 4685, miR-6718, miR-6733, miR-6808, and miR-8072) were cross-validated as being prognostic and retained for the age-adjusted panel. Panel-predicted survival estimates, grouped into tertiles, distinguished progressively worse survival at 2 years, OS was 81.0%+5.1% for upper tertile, 61.5%+6.3% for middle tertile, and 29.6%+6.3% for lower tertile. As shown in Table 1, each tertile included subjects at all levels of TNM stage, and tertile was not significantly associated with stage (chi-square test, p=0.26). Conclusions: This unique miRs panel represents a molecular staging system for gastric cancer that has the potential to prognosticate patients' survival more accurately than existing clinical systems can. [Table: see text]