New Algorithm Can ID Critical Cancer Mutations in DNA
July 12, 2022 – Most people probably know facial recognition as the thing that unlocks your smartphone. But this technology could also be used as a tool in the fight against cancer, according to a new study.
A team of researchers from University College London and the University of California, San Diego have developed an algorithm that works kind of like facial recognition – except instead of identifying faces, it picks out cancer mutations in DNA.
These mutations – what geneticists call “copy number changes” – are linked to different outcomes, some better and some worse, even among patients with the same cancer type.
“What’s been missing predominately in the field is a way to interpret those copy number changes,” says Nischalan Pillay, PhD, the University College London researcher who led the Nature study.
That’s what this algorithm does, Pillay says – it translates those changes into information that doctors could one day use to predict how a cancer is likely to behave. This may lead to more accurate outlooks, more effective treatments, and potentially more lives saved.
How Tech Can Find Cancer in DNA
Cancer is caused by DNA mutations, or, more simply put, “mistakes.” Some are tiny – like when just one letter of genomic code is off. These are “relatively easy to interpret,” Pillay says. But copy number changes are bigger. If your DNA s a book, copy number changes mean whole words, sentences, or entire pages can be wrong.
“It then becomes much harder to interpret,” Pillay says. “So, what we did was develop a way to summarize those, using patterns.”
To do that, he and his team analyzed nearly 10,000 cancer samples and discovered 21 cancer-related patterns. The algorithm can identify those patterns the way facial recognition software can find a suspect in a crowd.
For example: When facial recognition software finds a face, it breaks down all the parts – eyes, lips, nose, eyebrows – and uses them to build a digital version, comparing that to a database of known faces.
“It says, ‘OK, the closest similarity that this reconstructed face looks like is to X, Y, or Z person,’” says Pillay.
This algorithm finds not a face but a copy number change, breaking it down into each shattered, duplicated, or missing chromosome and making a profile that it can compare to those 21 known patterns, looking for a match.
“We’ve taken something that’s really complex and summarized that into a catalog, or a blueprint,” Pillay says.
That blueprint could be used to predict how a cancer is likely to progress, allowing doctors to closely monitor patients and try “a different form of therapy, or escalate the type of therapy,” depending on the patient’s chances of dying in a given time frame, says Pillay.
This Is Just the Beginning
Scientists are ever more interested in the role copy number changes may play in cancer treatment. For instance, these changes can also help show how a patient is likely to respond to a treatment, says Christopher Steele, PhD, a postdoctoral researcher at University College London and first author of the research.
Lab techs can already analyze copy number changes in blood samples, using liquid biopsies. As we learn more about how to interpret these results, doctors could use them to adjust treatment in real time, depending on how the cancer is evolving, Pillay says.
And someday, we may even come to understand how these copy number changes are caused in the first place, he says, possibly helping to prevent cancer.
It’s all part of an emerging subfield of cancer research that could revolutionize how we treat cancer.
“This is the very beginning,” Steele says.