Molecular features and predictive models identify the most lethal subtype and a therapeutic target for osteosarcoma

Area of Science:

• Oncology
• Molecular Biology
• Bioinformatics

Background:

• Osteosarcoma, the most common primary malignant bone tumor, has seen stagnant prognosis due to unchanged treatment over 30 years.
• Precise and personalized therapeutic strategies for osteosarcoma remain underexplored.

Purpose of the Study:

• To stratify osteosarcoma patients into molecular subtypes.
• To identify novel therapeutic targets and develop predictive models for osteosarcoma.

Main Methods:

• Non-negative matrix factorization (NMF) was used to stratify patients from discovery and validation cohorts.
• Survival analysis, transcriptomic profiling, and cell-based assays were employed.
• Machine learning tools (SVM, LASSO) were utilized to build predictive models.

Main Results:

• Four osteosarcoma subtypes (S-I to S-IV) were identified, with S-IV exhibiting the poorest prognosis and active cholesterol metabolism.
• SQLE, a key enzyme in cholesterol biosynthesis, was identified as a potential drug target for S-IV.
• Validated predictive models for subtype diagnosis and prognosis were developed using machine learning.

Conclusions:

• Molecular classification enhances understanding of osteosarcoma heterogeneity.
• Novel predictive models serve as robust prognostic biomarkers.
• Targeting SQLE presents a new therapeutic strategy for a subset of osteosarcoma patients.