AI Model Identifies Cancers of Unknown Primary with 95% Accuracy

Key Points

• AI model predicts cancer origins with 95% accuracy.
• Analyzes DNA methylation patterns from 7,500 patients.
• Could guide therapy development for cancers of unknown primary.

Full Details

Researchers from Kindai University in Japan developed a machine learning model that predicts cancer origins with 95% accuracy in test cohorts and 87% in validation cohorts. The model analyzes CpG-based DNA methylation patterns from nearly 7,500 patients with 21 cancer types. This approach could guide therapy development by identifying tumor biology patterns even after metastasis. The findings were presented at AACR 2026.