AI uncovers cancer-specific DNA methylation patterns

An AI model has been trained to recognize DNA methylation signatures of 13 different types of cancer.

Researchers at the University of Cambridge and Imperial College London (both UK) present an AI tool that could one day offer doctors an alternative cancer diagnostic method. The machine learning tool is trained to identify DNA methylation patterns associated with different cancer types, potentially allowing for earlier intervention.

DNA encodes genetic information that can be modified by environmental changes via the methylation of DNA bases – A, T, C and G. Each cell has millions of DNA methylation markers, which have been observed to change in early cancer development. By investigating the differences in DNA methylation between healthy and cancerous tissues using AI, the team hopes to identify methylation signatures of different cancer types.

Machine learning approaches for spatial omic data analysis

We caught up with Jeremy Goecks at AACR 2024 to discuss his lab’s work, get best practice tips for collecting and analyzing spatial omic data and discuss how the field can progress in the future.

The researchers combined machine and deep learning to develop their AI model, trained using tissue samples, rather than DNA fragments in blood. They stressed the importance of their model’s explainable and interpretable core, which provides reasoning behind its predictions. Although this model identified 13 cancers – including breast, lung, prostate and liver cancers – by their DNA methylation signatures with 98.2% accuracy, it requires additional training on a more diverse collection of biopsy samples before being used clinically.

“Computational methods such as this model, through better training on more varied data and rigorous testing in the clinic, will eventually provide AI models that can help doctors with early detection and screening of cancers,” concluded lead author Shamith Samarajiwa. “This will provide better patient outcomes.”