Background and Objective
Low-grade, stage Ta (TaLG) non-muscle-invasive bladder cancer (NMIBC) is generally nonaggressive; yet, some patients experience recurrence or progression. We previously identified long noncoding RNA cluster 2 (LC2) as an aggressive TaLG subgroup via consensus clustering, which required a cohort-level analysis. We developed a single-sample transcriptome classifier for patient-level LC2 identification.
Methods
Using the UROMOL cohort (n = 276) for training/testing and the Hurst (Stage-stratified molecular profiling of non-muscle-invasive bladder cancer enhances biological, clinical, and therapeutic insight. Cell Rep Med 2021;2:100472) cohort (n = 72) for validation, we performed feature selection via median absolute deviation and nested cross-validation. An elastic net regression model (α = 0.5) was trained using ten-fold cross-validation. Performance was evaluated through a pathway analysis, with prognostic significance assessed by Kaplan-Meier and univariable Cox regression.
KEY FINDINGS AND LIMITATIONS
Biological characterization revealed that LC2 tumors exhibited higher proliferation (G2M/E2F signatures), elevated FGFR3 pathway activity, and reduced sonic hedgehog signaling and immune activity (p < 0.001). The classifier identified 7/72 LC2 tumors in the Hurst cohort. LC2-TaLG patients had significantly worse recurrence-free survival (log-rank p < 0.001). On univariable Cox regression analysis, LC2 status was strongly associated with recurrence (hazard ratio 4.52, 95% confidence interval 1.78-11.5; p = 0.001). LC2-predicted tumors in the Hurst cohort showed similar biological patterns.
Conclusion
S AND CLINICAL IMPLICATIONS: We developed a single-sample transcriptomic classifier identifying aggressive TaLG NMIBC, demonstrating reproducibility across platforms. If validated, this model has the potential to support clinical decision-making.