Work

Machine learning-derived lysosome-dependent cell death signatures for prognostic stratification and immune microenvironment assessment in neuroblastoma

Year: 2026

Type: article

Abstract: Develop an LDCD-based prognostic model for neuroblastoma, identify core LDCD-related biomarkers, and characterize links to the immune microenvironment and therapeutic response. Bulk RNA-seq data from TARGET (n = 79) and GSE49710 (n = 483) were integrated; LDCD genes were curated from GeneCards. We evaluated 117 machine-learning model combinations (10 survival learners × 8 feature-selection strategies). Single-cell RNA-seq from GSE220946 (15 patients) was used to profile heterogeneity and cell–cell communication. Immune infiltration (ssGSEA), drug sensitivity to 198 compounds, and siRNA knockdown in five neuroblastoma cell lines were performed. ST6GAL2 and IL17RD were identified as LDCD-associated prognostic biomarkers that robustly stratified patients into risk groups with distinct survival. The optimized machine learning–derived risk model demonstrated consistent prognostic performance across training and validation cohorts, with area under the curve (AUC) values ranging from 0.715 to 0.833, exceeding a random baseline (AUC = 0.50). Importantly, this model complements established clinical systems, such as the INRG classification, by capturing LDCD-related transcriptional heterogeneity that is not explicitly represented in current clinicogenetic frameworks. Knockdown of ST6GAL2/IL17RD reduced cell viability across all tested cell lines. Single-cell analysis of 35,248 cells highlighted B cells as a prominent immune population linked to LDCD-related programs and metastatic features, and LDCD risk groups showed distinct immune infiltration, cell–cell communication, and drug sensitivity patterns. ST6GAL2 and IL17RD are prognostically relevant LDCD biomarkers in neuroblastoma, and the LDCD-informed model adds clinically complementary risk information while implicating B-cell–related signaling and lysosomal vulnerabilities as potential targets for refined risk-adapted therapy.

Source: Journal of Radiation Research and Applied Sciences

DOI: 10.1016/j.jrras.2026.102198

Authors: Shuran Zi, Di Yang, Yulei Hu

Institutions: Kunming University of Science and Technology, First People's Hospital of Yunnan Province

Language: English

Topic: Neuroblastoma Research and Treatments

Subfield: Neurology

Field: Medicine

Domain: Health Sciences

SDG: Good health and well-being

Open Access status: gold

Estimated APC paid: $1,400