Reviewed by Lexie CornerJun 20 2024
A new study in Biology Methods & Protocols, published by Oxford University Press, suggests that doctors may soon be able to use artificial intelligence (AI) to detect and diagnose cancer in patients, enabling earlier treatment. Investigators from Cambridge University and Imperial College London trained an AI model, using a combination of machine learning and deep learning, to analyze DNA methylation patterns and identify 13 different cancer types (including breast, liver, lung, and prostate cancers) from non-cancerous tissue with 98.2 % accuracy.
With over 19 million cases and 10 million deaths yearly, cancer continues to be one of the most challenging diseases to treat in humans. The treatment of late-stage tumors is difficult due to the evolutionary nature of cancer.
The patterns of the four bases—represented by the letters A, T, G, and C—that make up DNA's structure encode genetic information. Certain DNA bases may be altered by the addition of a methyl group due to environmental changes occurring outside the cell. This process is referred to as “DNA methylation.” There are millions of these DNA methylation marks in every single cell. These markers have been shown to alter during the early stages of cancer development, and this could help with early cancer diagnosis.
It is possible to investigate which DNA bases, and to what degree, are methylated in cancers as opposed to healthy tissue. Finding the distinct DNA methylation signatures corresponding to various cancer types is like finding a needle in a haystack. The study's researchers think AI can be useful in this situation.
To be ready for clinical use, this model would require more testing and training on a more varied collection of biopsy samples, as it is dependent on tissue samples rather than DNA fragments in the blood.
The use of an interpretable and explainable core AI model offered insights into the logic underlying the model's predictions, which the researchers deemed to be a significant component of this investigation. By delving into the inner workings of their model, the researchers were able to demonstrate how it improves and validates the understanding of the fundamental mechanisms causing cancer.
Medical professionals would be able to identify these uncommon methylation patterns (possibly from biopsies) and use them to identify cancer early. Most cancers are treatable or curable if discovered early enough, which could significantly improve patient outcomes.
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. This will provide better patient outcomes.
Shamith Samarajiwa, Study Lead Author, University of Cambridge
Journal Reference:
Newsham, I., et al. (2024) Early detection and diagnosis of cancer with interpretable machine learning to uncover cancer-specific DNA methylation patterns. Biology Methods and Protocols. doi.org/10.1093/biomethods/bpae028