Urothelial cancer is a challenging disease with a wide tumor-biological spec-trum. Most molecular classification attempts are transcriptome-based and re-late only indirectly to the therapeutically relevant protein level. We improve preanalytical preparation of clinical samples for proteome analyses and char-acterize a comprehensive cohort of 434 samples with 242 tumors and 192 paired normal mucosae. We evaluate sample-wise tumor specificity and rank biomarkers by target relevance. We identify five tumor clusters that add prog-nostic information independent from histopathological groups. In silico drug prediction suggests efficacy of several compounds hitherto not in clinical use. Both, in silico as well as in vitro data, indicate predictive value of the proteo-mic clusters for these drugs. Comparative validation on external transcriptom-ic data confirms prognostic relevance and contextualizes proteomic and tran-scriptomic classifications. A real-world diagnostic approach based on im-munohistochemistry further validates our proteomic data but underlines the necessity of omics scale molecular analyses for personalized oncology.