Comprehensive bladder cancer molecular profiling and monitoring based on mutational, epigenomic, and expression data from real-world patients.

Abstract

860 Background: Bladder cancer poses considerable therapeutic challenges, particularly in the context of bladder preservation strategies that aim to balance maintaining quality of life with achieving optimal survival outcomes. Traditional treatments, such as radical cystectomy, often have profound impact on patient morbidity. However, recent advances in molecular profiling -- including mutational, epigenomic, and gene expression analyses -- offer promising avenues for personalized cancer treatment approaches. Despite these advances, comprehensive studies leveraging detailed analysis of molecular profiling and urine-based monitoring in bladder cancer are lacking. Methods: A cohort of 30 real-world patients with bladder and upper tract urothelial carcinoma (UTUC) of various subtypes were included in this study. Standard treatment modalities employed were neoadjuvant chemotherapy, trimodality therapy (TMT) for bladder preservation, and systemic chemotherapy or immunotherapy in the metastatic setting. PredicineCOMPLETE is an integrated assay that interrogates genomic alterations such as mutations, copy number variations (CNVs), and gene fusions – as well as DNA methylation and transcriptomic profiles. All samples had previously undergone the PredicineBEACON urine MRD test, which covers mutation/CNV/fusion analysis. We further performed whole-transcriptome sequencing (WTS) on baseline urine cell pellet (UCP) samples, and the PredicineEPIC DNA methylation assay on all the follow-up timepoint samples. Results: The PredicineCOMPLETE assay generated significantly more comprehensive data compared to mutational panels alone, providing a rich molecular landscape of each patient’s tumor biology. Comprehensive correlation analysis between DNA methylation and RNA expression were conducted. Strong correlation was observed for some bladder cancer biomarker genes such as TERT and TP53. An AI model was trained using patients’ treatment decision (bladder preservation versus surgery), clinical features, and the comprehensive molecular data obtained. The trained AI model accurately matches the physician’s choices over 90% of the validation cases. Conclusions: This study demonstrates that comprehensive molecular profiling using the PredicineCOMPLETE assay helps us understand the underlying disease biology. Combined with machine learning and AI modeling, it can effectively assist in treatment decision-making for bladder cancer patients. While these initial results are encouraging, the limited sample size necessitates further validation in larger, independent cohorts to confirm the model’s efficacy and generalizability.