Predicting Cancer Metastasis Origins via CNV Data and Neural Networks: High Accuracy with Precision Trade-Off

Metastatic cancer is the spread of cancer from its original site to other parts of the body. In some cases, doctors cannot locate the place within the body from which the cancer first appeared, which is a crucial determination to be made for choosing the right treatment approach. Now, scientists have found a potential way out using DNA and artificial intelligence.

Every cell in our body has a unique DNA signature that is altered when cancer arises. These changes often come in the form of what is called “Copy Number Variations” or CNVs, sections of DNA that are duplicated or lost. These CNVs leave behind a genetic “fingerprint” pointing back to the original tumor.

In the research work, scientists have developed an AI tool, using a form of neural network capable of decoding such CNVs patterns. They then trained this AI on samples of over 18,000 tumors from 38 types of cancers. When the researchers tested this on patient samples with unknown primary cancers, it fared well by determining the site in most cases.

This is a game-changing approach: it uses a common test-a genomic test-already available in most laboratories; thus, this modality of treatment can easily be adopted by doctors in patient care. By knowing the origin of the cancer, doctors can determine more specific treatments, thus hopefully leading to better survival rates and a better quality of life.

The research is exemplifying an exciting step in the treatment of cancer. Using genetics and AI together to understand the unique cancer that an individual has opens a door to precision medicine: treatment designed exactly for what that specific individual needs.

Full text: Michel-Edwar Mickael, Norwin Kubick, Atanas G Atanasov, Petr Martinek, Jarosław Olav Horbańczuk, Nikko Floretes, Michael Michal, Tomas Vanecek, Justyna Paszkiewicz, Mariusz Sacharczuk, Piotr Religa, 2024, Using Copy Number Variation Data and Neural Networks to Predict Cancer Metastasis Origin Achieves High Area under the Curve Value with a Trade-Off in Precision, Current issues in molecular biology, 46(8), 8301–8319, https://doi.org/10.3390/cimb46080490.