Establish TIIC signature score based the machine learning fusion in bladder cancer

Bladder cancer is a prevalent malignant tumor with high heterogeneity. Current treatments, such as transurethral resection of bladder tumor (TURBT) and intravesical Bacillus Calmette-Guérin (BCG) therapy, still have limitations, with approximately 30% of non-muscle-invasive bladder cancer (NMIBC) progressing to muscle-invasive bladder cancer (MIBC), and a substantial number of MIBC patients experiencing recurrence after surgery. Immunotherapy has shown potential benefits, but accurate prediction of its prognostic effects remains challenging.

We analyzed bladder cancer RNA-seq data and clinical information from The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) databases, and used various machine learning algorithms to screen for feature RNAs related to tumor-infiltrating immune cells (TIICs) from single-cell data. Based on these RNAs, we established a TIIC signature score and evaluated its relationship with overall survival (OS) and immunotherapy response in bladder cancer patients.

The study identified 171 TIIC-RNAs and selected 11 TIIC-RNAs with prognostic value through survival analysis. The TIIC signature score established using a machine learning fusion method was significantly associated with OS and showed good predictive performance in different datasets. Additionally, the signature score was negatively correlated with immunotherapy response, with patients with low TIIC feature scores showing better survival outcomes after immunotherapy. Further biological functional analysis revealed a close association between the TIIC signature score and immune regulation processes, cellular metabolism, and genetic variations.

This study successfully constructed and validated an RNA signature scoring system based on tumor-infiltrating immune cell (TIIC) features, which can effectively predict OS and the effectiveness of immunotherapy in bladder cancer patients.