Abstract:Muscle-invasive bladder cancer (MIBC) is associated with poor predictability of response to cisplatin-based neoadjuvant chemotherapy (NAC). Consequently, the benefit of NAC remains unclear for many patients due to the lack of reliable biomarkers predicting treatment response. In order to identify biomarkers and build an integrated and highly accurate model to predict NAC response, we performed a comprehensive transcriptomic and genomic profiling on tumors from 100 MIBC patients. Our results showed that the expression of the top genes associated with response, as well as the expression of growth factor genes and cell cycle regulators are highly correlated with NAC response. Most importantly, we found a novel signature related to the WNT signaling pathway that alone was highly correlated with NAC response and showed high accuracy in predicting NAC response (AUC=0.76). Additionally, mutations in the DNAH family genes (DNAH8, DNAH6 and DNAH10) and deletion in KDM6A were also highly correlated with NAC response. Using our comprehensive molecular analysis as a backbone, we developed two machine learning (ML) models, one incorporating both transcriptomic and genomic features (RF-RW), and the other using only transcriptomic data (RF-R). Both models demonstrated promising performance (AUC=0.82) as predictive models of response to NAC in MIBC. RF-RW and RF-R, after external validation, could potentially change the management of MIBC patients by selecting ideal candidates for NAC.

What problem does this paper attempt to address?

This paper aims to solve the difficult problem of predicting the response of muscle - invasive bladder cancer (MIBC) patients to cisplatin - based neoadjuvant chemotherapy (NAC). Specifically, the main objective of the study is to characterize the molecular features of MIBC in terms of NAC response through comprehensive transcriptomics and genomics analysis. The secondary objective of the study is to develop a machine - learning model that can predict the treatment response of MIBC patients to NAC. The ultimate goal is to create a prediction tool that can be used in clinical practice to select ideal patients suitable for receiving NAC, thereby changing the management of MIBC. ### Main research background and problems - **Treatment status of MIBC**: MIBC is an aggressive cancer type with a poor prognosis. Approximately 20% of newly diagnosed bladder cancer patients are MIBC. The current standard treatment method is to use cisplatin - based neoadjuvant chemotherapy (NAC) before radical cystectomy. Although NAC has been proven to improve overall survival rate, less than 50% of patients are suitable for receiving NAC due to comorbidities or frailty, etc., and about 40% of patients do not respond to NAC. These patients have a high recurrence rate and a low 5 - year survival rate. - **Lack of reliable biomarkers**: Currently, there is a lack of reliable biomarkers to predict which patients will respond to NAC. This leads to the inability to accurately identify patients who may benefit from NAC. At the same time, it also makes non - responders unable to avoid unnecessary toxicity and miss other potentially more effective treatment options. ### Research methods - **Samples and data sources**: The study included 100 MIBC patients from four hospitals in the Catalonia region of Spain. All patients received cisplatin - based neoadjuvant chemotherapy. The research team collected tumor samples from these patients and performed RNA sequencing (RNA - Seq) and whole - exome sequencing (WES). - **Data analysis**: Differential expression analysis (DEA), weighted gene correlation network analysis (WGCNA) and pathway enrichment analysis were performed through RNA - Seq data to identify genes and signaling pathways related to NAC response. Variation detection (VC) and somatic copy number variation (SCNV) analysis were performed through WES data to explore the relationship between specific mutations and genomic features and NAC response. - **Machine - learning model**: Based on the above analysis results, the research team constructed two machine - learning models, one combined transcriptomics and genomics features (RF - RW), and the other only used transcriptomics data (RF - R), to predict the treatment response of NAC. ### Key findings - **WNT signaling pathway gene expression characteristics**: The study found that the expression characteristics of WNT signaling pathway - related genes were significantly related to NAC response. In particular, the expression characteristics of 17 WNT signaling pathway genes were significantly different between responders and non - responders. - **DNAH family gene mutations and KDM6A deletion**: The study also found that mutations in the DNAH8, DNAH6 and DNAH10 genes and deletion of the KDM6A gene were significantly related to NAC response. - **Molecular subtypes**: Although the relationship between the molecular subtypes of MIBC and NAC response was inconsistent, some subtypes (such as Luminal Papillary and Basal Squamous) showed a better trend of NAC response. ### Conclusion Through comprehensive transcriptomics and genomics analysis, this study successfully identified multiple biomarkers related to the NAC response of MIBC patients and developed two machine - learning models. These models showed high accuracy in predicting NAC response. These findings are expected to provide new tools for clinicians to more accurately select MIBC patients suitable for receiving NAC, thereby improving the treatment effect and quality of life of patients.

Predicting response to cisplatin-based neoadjuvant chemotherapy for muscle-invasive bladder cancer: transcriptomic features outrank genomic biomarkers