Identification of a gene expression signature for predicting Wilms’ tumor recurrence via RNA-seq and LASSO-Cox regression

Abstract

Wilms’ tumor (WT) is the most common form of pediatric kidney cancer, typically characterized by favorable outcomes but with a relatively high recurrence rate. The prognosis of recurrence often relies on histological features, tumor staging, and response to initial therapies. However, there is a need to incorporate molecular factors for more effective management of relapsed cases. In this study, we analyzed the transcriptomic profiles of WT using bulk RNA sequencing data obtained from The Cancer Genome Atlas. Gene expression profiles of recurrent tumors were compared with those of primary tumors, and primary tumors were compared to non-cancerous controls to identify dysregulation in gene expression programs. Our results revealed that recurrent WT exhibit transcriptional profiles distinct from those of primary tumors. While a subset of dysregulated genes was shared between primary and recurrent WT, several genes were uniquely differentially expressed in recurrent tumors. Functional enrichment analyses showed that these expression differences primarily affect immune response, cell adhesion, and extracellular matrix components. Using Least Absolute Shrinkage and Selection Operator (LASSO) regression together with Cox and Kaplan-Meier analyses, a 20-gene risk signature was established, which is strongly associated with relapse probability. The model achieved its highest time-dependent area under the receiver-operating characteristic curve of 91.5% when predicting five-year recurrence survival. Our analyses also demonstrated that, along with the risk score, male gender and the favorable histological subtype of primary tumors are also linked to poorer disease-free survival. These findings serve as preliminary evidence toward establishing gene expression signatures that could aid risk stratification of WT.