Unraveling the PANoptosis Landscape in Osteosarcoma: A Single-Cell Sequencing and Machine Learning Approach to...

Background: Osteosarcoma (OS) is a highly aggressive bone malignancy prevalent in children and adolescents, characterized by poor prognosis and limited therapeutic options. The tumor microenvironment (TME) and cell death mechanisms such as PANoptosis-comprising pyroptosis, apoptosis, and necroptosis-play critical roles in tumor progression and immune evasion. This study is aimed at exploring the PANoptosis landscape in OS using single-cell RNA sequencing (scRNA-seq) and at developing a robust prognostic model using machine learning algorithms. Methods: Single-cell sequencing data for OS were obtained from the GEO database (GSE162454), and bulk transcriptome data were sourced from the TARGET and GEO databases. Data integration, dimensionality reduction, and cell clustering were performed using UMAP and t-SNE. PANoptosis-related genes were identified, and their expression profiles were used to score and categorize cells into PANoptosis-high and PANoptosis-low groups. A comprehensive prognostic model was constructed using 101 machine learning algorithms, including CoxBoost, to predict patient outcomes. The model's performance was validated across multiple cohorts, and its association with the mutation landscape and TME was evaluated. Results: The scRNA-seq analysis revealed 14 distinct cell clusters within OS, with significant PANoptosis activation observed in cancer-associated fibroblasts (CAFs), myeloid cells, osteoblasts, and osteoclasts. Differentially expressed genes between PANoptosis-high and PANoptosis-low groups were identified, and cell communication analysis showed enhanced interaction patterns in the PANoptosis-high group. The CoxBoost model, selected from 101 machine learning algorithms, exhibited stable prognostic performance across the TARGET and GEO cohorts, effectively stratifying patients into high-risk and low-risk groups. The high-risk group displayed worse survival outcomes, higher mutation frequencies, and distinct immune infiltration patterns, correlating with poorer prognosis and increased tumor purity. Conclusion: This study provides novel insights into the PANoptosis landscape in OS and presents a validated prognostic model for risk stratification. The integration of scRNA-seq data with machine learning approaches enhances our understanding of OS heterogeneity and its impact on patient prognosis, offering potential avenues for targeted therapeutic strategies. Further validation in clinical settings is warranted to confirm the model's utility in guiding personalized treatment for OS patients.