Image: Microscopic image of prostate cancer.
Our genetic information is encoded in our DNA by the four base patterns A, T, G and C. Changes in the extracellular environment can modify some of our DNA bases by adding a methyl group to them, which is called DNA methylation.
Researchers have observed changes in these methylations in the early development of cancer. This can help in early diagnosis, because it is possible to check which bases in the DNA are methylated in cancer, and which are not present or not to the same extent in healthy tissue. But identifying specific DNA methylation patterns that distinguish different types of cancer is like searching for a needle in a haystack. Artificial intelligence can help with this.
Early stage
Researchers from the University of Cambridge and Imperial College London have trained an artificial intelligence system that uses machine learning and deep learning to look at patterns of DNA methylation. They were able to identify thirteen different types of cancer (including breast, liver, lung, and prostate) in tissue samples with 98.2 percent accuracy.
The model needs further training on a more diverse set of biopsy samples and more comprehensive testing before it is ready to help healthcare providers detect cancer early. This will greatly improve patients’ chances, as most cancers are treatable and often curable if detected early enough. The evolutionary nature of cancer makes it difficult to treat late-stage tumors.
According to the researchers, an important aspect of their research was the use of an interpretable and explainable AI model, which provided insight into the reasons behind the predictions. The researchers showed that the model enhances and improves understanding of the basic processes that contribute to cancer.