Skip to main content
CTOS 2024 Main banner
Final Program


Schedule at a Glance
Icon Legend
This session is not in your schedule.
This session is in your schedule. Click again to remove it.
Back

(P 83) PREDICTION OF HIGH-RISK OSTEOSARCOMA PATIENTS BASED ON MACHINE LEARNING USING TRANSCRIPTOME AND METHYLATION DATA FROM SGH-COHORT

    Co-Author(s)

  • WZ

    Weisong Zhao, n/a

    Department of Orthopaedics, Shanghai General Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China; Shanghai Bone Tumor Institution, Shanghai, China
    Shanghai, Shanghai, China (People's Republic)

  • ZD

    Zhenwu Dai, n/a

    Jiangsu Simcere Diagnostics Co., Ltd., Nanjing Simcere Medical Laboratory Science Co.
    Shanghai, Shanghai, China (People's Republic)

  • Yafei Jiang, MD

    85 Wujin Road. Shanghai, China.
    Department of Orthopedic Oncology, Shanghai General Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China
    Shanghai, Shanghai, China (People's Republic)

  • YS

    Yunjie Song, n/a

    Jiangsu Simcere Diagnostics Co., Ltd., Nanjing Simcere Medical Laboratory Science Co.
    Nanjing, Jiangsu, China (People's Republic)

  • KT

    Kai Tian, n/a

    Department of Orthopaedics, Shanghai General Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China; Shanghai Bone Tumor Institution, Shanghai, China
    Shanghai, Shanghai, China (People's Republic)

  • GW

    Gangyang Wang, PhD, MD

    Department of Orthopaedics, Shanghai General Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China; Shanghai Bone Tumor Institution, Shanghai, China
    Shanghai, Shanghai, China (People's Republic)

  • WS

    Wei Sun, MD

    Department of Orthopaedics, Shanghai General Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China; Shanghai Bone Tumor Institution, Shanghai, China
    Shanghai, Shanghai, China (People's Republic)

  • zc

    zhengdong cai, M.D.

    Shanghai General Hospital, School of Medicine Shanghai Jiao Tong University
    Shanghai, Shanghai, China (People's Republic)

    • Author(s)

  • Yingqi Hua

    Professor
    Department of Orthopedic Oncology, Shanghai General Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China
    Shanghai, Shanghai, China (People's Republic)

Objective: Osteosarcoma (OS) is the most prevalent primary malignant bone sarcoma, characterized by its high rates of metastasis and mortality. Our previous multi-omics study identified four OS subtypes with distinct molecular features and clinical outcomes. Among these subtypes, MYC-driven (S-MD) subtype showed the poorest prognosis and was termed high-risk OS. The most significant prognostic role of high MYC in OS for poor outcome has also been validated by other groups. Osteosarcoma should no longer be considered as a single disease since MYC status is relevant to high risk. Available diagnostic tool is needed for clinicians to identify high-risk OS in advance. The aim of this study is to develop a classifier for predicting the high-risk OS subtype.

Methods: In this study, we utilized the eXtreme Gradient Boosting (XGBoost) algorithm with Bayesian optimization to develop a classification model by integrating transcriptome and methylome data from our internal cohort (SGH-OS cohort). Furthermore, an external cohort (TARGET-OS cohort) was used to validate the classifier's predictive performance.

Results: Using the XGBoost algorithm, we developed a fine-tuned model consisting of nine risk-associated genes (ARHGAP9, CADM1, CPE, DUSP3, FGFR1, GALNT3, IGF2BP3, KIF26A, ZFP3). Within our internal cohort, the classifier demonstrated excellent predictive performance, achieving a high area under the ROC curve (AUC) of 0.999 and an overall accuracy of 0.989. Moreover, the classifier effectively stratified two groups with distinct survival patterns in the TARGET-OS cohort.

Conclusion: Our classifier exhibited excellent predictive performance in identifying high-risk OS patients, potentially enhancing treatment decision-making and optimizing patient management strategies.