Immune checkpoint blockade has shown limited efficacy in bladder cancer, underscoring the need for therapeutic strategies that enhance tumor immunogenicity. Immunogenic cell death (ICD) can stimulate adaptive immunity and remodel the tumor microenvironment, yet its comprehensive landscape and druggable potential in bladder cancer remain unexplored. Here, we performed integrative bioinformatics analysis of transcriptomic data from TCGA and GEO bladder cancer cohorts to characterize ICD-associated subtypes, construct a prognostic risk model, and analyze immune infiltration. Connectivity Map (CMap) database and the XSum algorithm were employed for computational drug screening. Molecular docking was used to predict interactions between candidate compounds and immune checkpoint proteins. In silico ADMET and toxicology profiles were evaluated. The clinical relevance was assessed through independent CT urography imaging and PD-L1 immunohistochemistry data. Two ICD molecular subtypes were identified with distinct prognostic and microenvironment features. An eight-gene ICD risk model was established and validated, effectively stratifying patients into high- and low-risk groups with differential survival, immune cell infiltration, and tumor mutational burden. Through computational screening, NU-1025 was nominated as a top candidate capable of reversing the high-risk gene expression pattern. In silico profiling indicated favorable drug-like properties and low toxicity risk. Clinical data analysis revealed that high-grade bladder tumors frequently exhibit low PD-L1 expression, aligning with the predicted need for ICD-inducing agents. Overall, this research decodes the ICD landscape in bladder cancer and, via a computational drug repurposing pipeline, nominates NU-1025 as a novel candidate with predicted ICD-inducing and/or immune checkpoint modulatory activity.
High-throughput zebrafish screening generates large volumes of imaging data, creating a need for scalable computational methods that reduce dependence on manual annotation and repeated task-specific model development. In this study, we develop a self-supervised foundation model for zebrafish image analysis pretrained on more than 468,000 images from seven heterogeneous zebrafish imaging datasets spanning different developmental stages, imaging platforms, modalities, and experimental conditions. By learning a shared representation from diverse unlabeled zebrafish images, the proposed framework provides a reusable encoder that can be adapted to multiple biological endpoints and annotation types, reducing the need to learn independent representations for each downstream task. We evaluated two self-supervised pretraining strategies: Self-Distillation with No Labels (DINO) and a sequential Masked Autoencoder (MAE) followed by DINO (MAE+DINO), across morphology classification, view classification, developmental stage prediction, fish localization, embryo clustering, and anatomical segmentation. The learned representations captured visual similarity, developmental progression, embryo-specific trajectories, phenotypic information, and anatomical structure across heterogeneous imaging data. MAE+DINO achieved the strongest overall performance for morphology classification, view classification, developmental stage prediction, and fine-tuned localization, while DINO provided the strongest frozen representation for localization. For anatomical segmentation, zebrafish specific pretraining improved performance when the encoder was frozen, whereas MAE+DINO and ImageNet initialization achieved comparable overall performance after end-to-end fine-tuning with a SegFormer-style decoder. These results demonstrate that organism-centered self-supervised pretraining can provide reusable and biologically informative visual representations that support diverse zebrafish image-analysis tasks and offer a scalable, annotation-efficient alternative to repeatedly developing task-specific representations.
