Integrating transcriptomics, single-cell omics, and deep learning-based histopathological features to identify OLFML3 as in bladder cancer

Background Bladder cancer (BCa) has the highest recurrence rate among solid tumors (61% at 1-year). Suboptimal surveillance and heterogeneity-limited models hinder precision. This study integrates multi-omics data to reveal the critical role of OLFML3 in BCa recurrence. A deep learning-based predictive model using pathology features offers personalized monitoring for high-risk patients. Methods Based on RNA sequencing data and clinical information from TCGA, BCa patients were divided into relapse and non-relapse groups. Weighted Gene Co-expression Network Analysis(WGCNA) identified gene modules associated with one-year relapse, and univariate Cox regression and LASSO Cox regression were used to select eight prognostic genes. A risk model was developed and validated in TCGA and GEO datasets. Independent survival analysis and clinical variable associations were performed for these genes. Single-cell RNA sequencing from the Guangdong Provincial Second People's Hospital cohort was used to analyze gene expression in different BCa subtypes. A deep learning model was developed to predict OLFML3 expression in H&E-stained images. Statistical analysis was performed using R software (version 4.4.2), with significance set at P < 0.05. Results WGCNA identified gene modules associated with BCa recurrence, with the red module showing a significant positive correlation. Univariate and LASSO Cox regression analyses selected 8 prognostic genes. Kaplan-Meier analysis revealed that patients with high OLFML3 expression had lower survival rates in both the TCGA and GEO datasets. High OLFML3 expression correlated with BCa invasiveness, grade, and stage. Single-cell RNA sequencing revealed differential expression of OLFML3 between high and low tumor-stroma subtypes, with OLFML3 identified as a key gene associated with 1-year BCa recurrence. A random forest model successfully predicted OLFML3 expression across different datasets. Conclusions Multi-omics approaches effectively identified the OLFML3 gene as a key gene associated with 1-year BCa recurrence and successfully constructed a deep learning model using pathological features to predict OLFML3 expression. Further research is needed to clarify its role in the progression of 1-year BCa recurrence.