Identification of the shared gene signatures in retinoblastoma and osteosarcoma by machine learning

Osteosarcoma (OS) is the most prevalent secondary sarcoma associated with retinoblastoma (RB). However, the molecular mechanisms driving the interactions between these two diseases remain incompletely understood. This study aims to explore the transcriptomic commonalities and molecular pathways shared by RB and OS, and to identify biomarkers that predict OS prognosis effectively. RNA sequences and patient information for OS and RB were obtained from the University of California Santa Cruz (UCSC) Xena and Gene Expression Omnibus databases. When RB and OS were first identified, a common gene expression profile was discovered. Weighted Gene Co-expression Network Analysis (WGCNA) revealed co-expression networks associated with OS after immunotyping patients. To evaluate the genes shared by RB and OS, univariate and multivariate Cox regression analysis were then carried out. Three machine learning methods were used to pick key genes, and risk models were created and verified. Next, medications that target independent prognostic genes were found using the Cellminer database. The comparison of differential gene expression between OS and RB revealed 1216 genes, primarily linked to the activation and proliferation of immune cells. WGCNA identified 12 modules related to OS immunotyping, with the grey module showing a strong correlation with the immune-inflamed phenotype. This module intersected with differential genes from RB, producing 65 RB-associated OS Immune-inflamed Genes (ROIGs). Analysis identified 6 hub genes for model construction through univariate Cox regression and three machine learning techniques. A risk model based on these hub genes was established, demonstrating significant prognostic value for OS. Genes shared between OS and RB contribute to the progression of both cancers through multiple pathways. The ROIGs risk score model independently predicts the overall survival of OS patients. Additionally, this study highlights genes with potential as therapeutic targets or biomarkers for clinical use.