Single-cell analysis reveals that NRG1/3–ERBB4 signaling affects metabolic reprogramming and immune escape in Wilms tumor

Wilms tumor presents a heterogeneous tumor microenvironment. This study aimed to characterize the tumor microenvironment and identify prognostic genes to improve therapeutic strategies. We integrated single-cell RNA sequencing and transcriptomic data from paired tumor and normal tissues. Bioinformatics analyses included assessments of cellular heterogeneity, trajectories, and cell–cell communication. Prognostic genes were identified with differential expression, Cox regression, and machine-learning analyses. Furthermore, functional characterization, immune infiltration patterns, and therapeutic targets were systematically investigated. Potential therapeutic compounds were predicted using drug databases and a graph-based deep learning framework to predict compound–protein interactions. Single-cell RNA sequencing revealed 17 cell clusters, with tumor-specific epithelial cells and renal progenitor cells. Pseudotime trajectory analysis revealed dynamic differentiation, highlighting NRG1/3–ERBB4 signaling. Intersection of the transcriptomic and single-cell data identified 405 key genes. A prognostic model incorporating eight prognostic genes (PRLR, SLC16A7, SGIP1, PPARGC1A, CDHR5, GRB7, FKBP10, and UGT2B7) stratified patients into high- and low-risk groups (p < 0.0001), with area under the curve values > 0.6 for 1-, 2-, and 3-year survival prediction. High-risk patients had elevated regulatory T cell infiltration and immune checkpoint genes (TNFRSF9 and KIR3DL3). Chitosan was identified as a multitarget agent that interacts with the eight prognostic proteins. This study defined tumor cellular architecture and identified eight prognostic genes with potential clinical value.