Integrative single-cell and bulk transcriptomics identify an <i>ALK</i>-associated three-gene signature predicting neuroblastoma outcomes

<h4>Background</h4>Activating anaplastic lymphoma kinase (<i>ALK</i>) mutations define a clinically relevant subset of neuroblastoma (NB), yet <i>ALK</i>-linked transcriptional programs with robust prognostic value remain insufficiently characterized. This study aimed to identify <i>ALK</i>-associated transcriptional features across single-cell and bulk datasets and to develop a reproducible prognostic signature for NB.<h4>Methods</h4>We integrated a public single-cell atlas of high-risk NB from CELL×GENE with bulk RNA sequencing (RNA-seq) cohorts and NB cell-line datasets. Differential expression between <i>ALK</i>-mutant and <i>ALK</i>-wild-type samples was analyzed in (I) pseudobulk malignant cells from the single-cell dataset, (II) the Kids First (Maris) cohort, and (III) GSE89413 NB cell lines, followed by direction-consistent intersection. A prognostic model was developed in GSE62564 using Cox regression and evaluated by Kaplan-Meier survival analysis and time-dependent receiver operating characteristic (ROC) curves. External validation was performed in GSE181559 and TARGET-NBL. Biological significance was further explored using preranked gene set enrichment analysis (GSEA).<h4>Results</h4>Three concordant <i>ALK</i>-associated genes, <i>DLK1</i>, <i>RTL1</i>, and <i>ISLR2</i>, were identified across all three analytical contexts and were consistently upregulated in <i>ALK</i>-mutant samples. A three-gene risk score significantly stratified overall survival and event-free survival in GSE62564 and remained predictive in GSE181559 and TARGET-NBL without model re-fitting. High-risk tumors were enriched for translational and ribosome biogenesis pathways, whereas low-risk tumors were enriched for adaptive immune activation programs.<h4>Conclusions</h4>Cross-context transcriptomic integration identified a minimal <i>ALK</i>-associated three-gene signature that reproducibly predicts NB outcomes and reflects distinct underlying biological states, supporting its potential value for risk stratification and hypothesis-driven therapeutic development.