Development and Validation of a Novel liquid-liquid phase separation Gene Signature for Bladder Cancer

Abstract

Abstract Background Bladder carcinoma (BLCA) is a prevalent malignancy in the urinary tract and is known for its aggressive nature and high probability of recurrence. The regulation of the biological response during tumor proliferation and metastasis is inextricably linked to liquid-liquid phase separation (LLPS). To facilitate early detection and treatment, this study utilized transcriptomic data to explore the prognostic roles of LLPS-linked genes and develop a predictive model.Methods The dataset of bladder cancer patients consisted of clinical and transcriptomic data retrieved from the GEO and TCGA databases. The study utilized a clustering algorithm that employed non-negative matrix factorization (NMF) to classify the samples, which were further compared systematically for their liquid-liquid phase separation characteristics. A multivariate Cox regression model and the Least Absolute Shrinkage Selection Operator (LASSO) algorithm were utilized to construct prognostic models to establish risk formulas for nine genes. Validation of the gene signature was conducted in the entire TCGA cohort (406 cases), TCGA testing cohort (120 cases), and the external validation dataset GSE13507. The signature system was evaluated using both receiver operating characteristic (ROC) and Kaplan-Meier curves. Furthermore, decision curve analysis including the clinicopathological parameters and genetic signature was utilized to predict individual survival.Results In the study, two distinct molecular subtypes were identified, namely C1 and C2. It was revealed that individuals with the C1 subtype had a significantly more favorable prognosis as compared to those with the C2 subtype. Patients belonging to the low-risk group had remarkably better prognoses compared to those in the high-risk groups across the entire TCGA, GEO, and TCGA training cohorts. In addition, the LLPS-related gene model constructed in the study was validated as a prognostic factor independent of other clinical traits.Conclusions This study identifies gene clusters associated with LLPS and establishes a model that can predict, accurately and independently, the prognosis of BLCA. This model can be utilized in clinical practice to assess the prognosis of BLCA.