Synergistic machine learning models utilizing ferroptosis-related genes for improved neuroblastoma outcome prediction

Neuroblastoma (NB) is a highly heterogeneous and common pediatric malignancy with a poor prognosis. Ferroptosis, an iron-dependent cell death pathway, may play a crucial role in NB tumor progression and immune response. This study aimed to investigate ferroptosis in NB to identify potential therapeutic targets and develop predictive models for prognosis and recurrence.

Six datasets were accessed from the ArrayExpress database and Gene Expression Omnibus. Ferroptosis-related genes (FRGs) were selected from the FerrDb website. Unsupervised clustering, differential expression analysis, weighted correlation network analysis (WGCNA), and gene set enrichment analysis (GSEA) were adopted to investigate potential pathways associated with ferroptosis in NB and identify the key genes involved. We used the least absolute shrinkage and selection operator (LASSO) and multivariate Cox regression to develop the ferroptosis-related prognostic signatures (FRPS) while using machine learning (ML) algorithms to construct the recurrence model.

Ribosome and cell cycle may be the potential pathways for ferroptosis involved in NB, with MYCN and RRM2 identified as key genes in this regulatory process. Five FRGs-ATG7 (-1.009), ELAVL1 (1.739), PPARA (0.493), RDX6 (1.457), and TERT (0.247)-were screed out for the FRPS, which showed excellent predictive performance in comparison with other published NB signatures. Eight FRGs-ALDH3A2 (48.597), TERT (23.398), ULK2 (21.034), AKR1C1 (20.699), MFN2 (12.575), SLC16A1 (12.342), TF (10.240), and DDR2 (7.598)-were selected based on the importance scores to construct the recurrence model. Among the models, utilizing random forest (RF), XGboost, support vector machine (SVM), K-nearest neighbors (KNN), and linear discriminant analysis (LDA), the RF model exhibited the highest performance.

We investigated the potential ferroptosis-related pathways and hub- FRGs in NB and developed prognosis and recurrence models, providing new potential targets for prognostic evaluation and treatment in NB patients.