A support vector machine classifier for the prediction of osteosarcoma metastasis with high accuracy

Area of Science:

• Oncology
• Bioinformatics
• Genomics

Background:

• Osteosarcoma (OS) metastasis is a major challenge in treatment.
• Identifying genes driving OS metastasis is crucial for improved diagnostics and therapies.

Purpose of the Study:

• To identify critical genes associated with osteosarcoma metastasis.
• To construct and validate a predictive classifier for OS metastasis using gene expression data.

Main Methods:

• Meta-analysis of gene expression datasets (GEO) to identify differentially expressed genes (DEGs).
• Construction of a protein-protein interaction (PPI) network and identification of hub genes using betweenness centrality (BC).
• Development and validation of a Support Vector Machine (SVM) classifier using selected hub genes.

Main Results:

• 353 DEGs were identified between metastatic and non-metastatic OS.
• A PPI network revealed 64 key genes, which were used to build an SVM classifier.
• The SVM classifier achieved high prediction accuracies (up to 100%) across four independent datasets.
• Enrichment analysis highlighted pathways like 'pathways in cancer' and 'TGF-β signaling pathway' related to OS metastasis.

Conclusions:

• A robust SVM classifier for predicting OS metastasis was developed and validated.
• Key genes and pathways involved in osteosarcoma metastasis were identified.
• These findings offer potential for novel genetic diagnostic methods and a deeper understanding of OS metastasis mechanisms.