SAN DIEGO – Researchers from Kindai University in Japan have developed a machine learning model that accurately predicts the origin of diverse cancer types in patients with cancers of unknown primary (CUP) by analyzing CpG-based DNA methylation. Results showed that the model correctly identified the cancer type in about 95% of cases in the test cohort, and achieved 87% accuracy when applied to an independent validation cohort from 31 cases representing 17 different cancer types. The work was presented at the American Association for Cancer Research (AACR) Annual Meeting.  

“Our findings suggest that DNA-based approaches can help identify where a cancer may have started, even when the original tumor is not visible,” said Marco A. De Velasco, PhD, a faculty member in the department of genome biology at Kindai University in Japan.  

CUP are metastatic malignancies in which the primary cancer site could not be identified. These cancers are often associated with poorer outcomes, as patients are typically treated with broad, nonspecific chemotherapy regimens rather than therapies targeted to a specific cancer type. 

Approximately only 15-20% of patients with CUP show features that allow site-specific therapies. Patients receiving site-directed therapy can survive up to 24 months, compared with six to nine months for those receiving standard treatment. 

Patterns in tumor biology, such as gene activity or chemical modifications to DNA, can differ between cancer types and persist even after the cancer has spread and guide development of these therapies. While some methods have shown promise, they have yet to demonstrate clear survival benefits in clinical trials. 

The model was developed using methylation data from nearly 7,500 patients with 21 different cancer types obtained from The Cancer Genome Atlas Program and other public datasets. Using machine learning, the researchers identified CpG methylation and built methylation profiles that were associated with different tumor types. 

Del Velasco emphasized that the study achieved high accuracy in predicting the origin of diverse cancer types using a small subset of DNA markers, about 1,000 CpG regions selected from hundreds of thousands across the genome. “This is important because it shows that we can simplify complex molecular data while still maintaining strong predictive performance,” he said. 

As a limitation, the model was developed using cancers with known origins, rather than true CUP. Testing in CUP patients is important to understand how well the model performs in clinical settings. Additionally, not all tumors are easily accessible for genetic testing, particularly tumors in advanced stage. Looking ahead, the authors aim to adapt and evaluate the model using blood-based biopsy to analyze circulating tumor DNA instead of relying on DNA from tissue samples. 

The post AACR 2026: Cancers of Unknown Primary Identified by DNA Methylation AI Model appeared first on GEN – Genetic Engineering and Biotechnology News.