Comprehensive multi-layered omics analysis of the UBC landscape for precision medicine and its clinical implications: From deciphering distinct molecular heterogeneity to guiding treatment decision-making and developing tailored therapeutic strategies.

840 Background: Conventional transcriptomic subtyping of non-muscle invasive (NMIBC) and muscle-invasive bladder cancer (MIBC) fails to capture the complex molecular heterogeneity of urothelial bladder cancer (UBC). Our goal is to apply high-resolution multi-omics profiling to optimize risk stratification and management, and to elucidate novel biomarkers and therapeutic targets. Methods: Comprehensive multi-omics analyses were performed on 212 fresh frozen tumor tissues from 168 UBC patients (130 NMIBC and 38 MIBC) who underwent surgery at the National Cancer Center between 2020 and 2023. Results: Overall, the median mRNA–protein correlation (q = 0.307) was lower than that in other cancer types. Given the discordance, we integrated RNA sequencing, global proteomics, and phosphoproteomics for unsupervised clustering with non-negative matrix factorization (NMF). We identified four multi-omic clusters (NMF1–4) that encompass both NMIBC and MIBC, reflecting diverse oncogenic pathways and interactions within the tumor microenvironment. A Sankey diagram revealed partial overlaps between multi-omics NMF clusters and conventional transcriptomic subtypes. Furthermore, the limited and unpredictable efficacy of intravesical BCG and FGFR3 inhibition underscores the need for novel biomarkers and co-targets via a multi-omics approach. Genomic alteration analysis revealed that FGFR3 showed the strongest cis-effects, elevating both mRNA and protein levels. At the trans-acting level, FGFR3 alterations were linked to enhanced glycolytic processes. Proteogenomic analysis revealed FGFR3-mutant tumors enriched in extracellular matrix (ECM) and glycoprotein pathways with suppressed immune-regulatory programs. RNA-seq deconvolution and integrated transcriptomic–proteomic analyses revealed ECM remodeling, cancer-associated fibroblast (CAF) signatures, and the ECM modulator PLOD1 upregulation with reduced CD8⁺ T-cell infiltration in BCG non-responders. A strong correlation between PLOD1 ex-pression and CAF abundance in BCG non-responders, along with elevated SPP1 phosphorylation in the CAF(high)/PLOD1(high) subset suggests that disrupting the CAF–PLOD1–SPP1 circuit alongside BCG treatment may effectively address BCG failure, based on a novel framework linking CAF activation, ECM remodeling, and phospho-signaling dynamics. Conclusions: Our integrative multi-omics approach identified clinically relevant clusters that stratify all stages of UBC and bridged genomic–transcriptomic heterogeneity to actionable protein targets for precision oncology. Future efforts should focus on translating these findings into clinical applications while overcoming widespread adoption challenges.