Machine Learning Pathomics Model Predicts NRG1 Expression in Bladder Cancer

A machine learning-based pathomics model demonstrated moderate discriminatory ability in predicting NRG1 expression and its molecular-pathological features in bladder cancer. The study aimed to predict the expression of NRG1 in bladder cancer using pathomic techniques.

To construct and evaluate the prediction framework, the researchers gathered transcriptomic data, pathological images, and clinical data of bladder cancer patients from The Cancer Genome Atlas. The OTSU algorithm was applied to carry out image segmentation across the pathological images, preparing the tissue features for subsequent computational evaluation and feature selection.

Feature selection was conducted using mRMR_RFE screening, which identified a total of 20 relevant features from the segmented pathological data. These 20 selected features were subsequently modeled using a Gradient Boosting Machine algorithm to build the NRG1 prediction model. Evaluation of the receiver operating characteristic curves demonstrated an area under the curve of 0.789 in the training cohort and 0.702 in the validation cohort.

Beyond evaluating predictive accuracy, the researchers analyzed the relationship between the computed Pathomics score and immune profiles, finding that the Pathomics score was positively correlated with immune-related genes (p < 0.05). Furthermore, in bladder cancer cell assays, experimental silencing of NRG1 suppressed cell proliferation, migration, and invasion, whereas overexpression of NRG1 exhibited contrasting effects on these cellular activities.

Subject of Research: Machine learning prediction of NRG1 expression in bladder cancer

Article Title: Predicting NRG1 expression and its molecular-pathological features of bladder cancer through a machine learning-based pathomics model

Article References: Zhang, C., Wu, F., Zhou, Q., Duan, Q., Zhu, L., & Huang, W. (2026). Predicting NRG1 expression and its molecular-pathological features of bladder cancer through a machine learning-based pathomics model. Discover Oncology. https://doi.org/10.1007/s12672-026-06230-3

Image Credits: AI Generated

DOI: 10.1007/s12672-026-06230-3

Keywords: NRG1, bladder cancer, BLCA, pathomics, machine learning, GBM algorithm, OTSU algorithm, mRMR_RFE, TCGA, transcriptomic data, pathological images, ROC curve, AUC, cell proliferation, cell migration, cell invasion, immune-related genes, Pathomics score