<p>Uveal melanoma (UVM) is an aggressive intraocular malignancy with a high risk of metastasis but few effective therapies. However, current molecular classification systems do not completely reflect the immune-stromal interactions underlying tumor progression and immune dysfunction. Therefore, we aimed to establish a multidimensional immune ecological framework for UVM by integrating bulk and single-cell transcriptomics with analyses of immune activity, metabolic programs, regulon states, and cell–cell communication. We identified two robust immune ecological subtypes, CS1 and CS2, with markedly divergent clinical outcomes. The poor-prognosis CS1 subtype was characterized by frequent <i>BAP1</i> loss, prominent IFN/STAT signaling, and a paradoxical “inflamed yet immunosuppressed” tumor microenvironment, whereas CS2 was characterized by a relatively immune-stable reference state. Through machine-learning screening across independent cohorts, we identified the transcription factor RUNX1 as a prognostic marker enriched in the CS1 subtype. Functional assays demonstrated that RUNX1 promotes UVM cell proliferation and migration, suggesting a potential role in CS1 immune ecological state. Given its enrichment in CS1, RUNX1 may be related to cytokine signaling and stromal remodeling features; however, this inference is based on subtype characteristics and requires further validation. Consistent with these subtype-specific features, CS1 tumors exhibited distinct sensitivities to dasatinib, lapatinib, paclitaxel, and cisplatin, indicating immune-state-associated therapeutic vulnerabilities. Together, these findings suggest a RUNX1-associated immunosuppressive tumor ecology in UVM and provide a conceptual framework for immune-state-guided therapeutic strategies.</p>