Cancer is often found when someone starts to feel symptoms—pain, an abnormal growth, or maybe just fatigue. Now scientists have developed a computer program that could identify and locate cancer even before symptoms arise, opening the door for earlier screening and better treatment.The program, called CancerLocator, detects tumor DNA in patient blood samples, and precisely pinpoints where the tumor is located in the body. In a small pilot study, it successfully diagnosed liver, lung, and breast cancer in 80 percent of cases, giving researchers hope that the program could eventually be used as part of regular health checks, eliminating the need for invasive biopsies.
Developed by researchers from the University of California, Los Angeles, and the University of Southern California, CancerLocator works by analyzing DNA that escapes into the bloodstream when cells die. Each fragment of DNA has a unique pattern of chemical add-ons, called methyl groups, that mark which genes were turned on or off. These methyl markers can indicate whether a gene was interrupted in a cancerous cell. And because different cells and tissues have different genes that are activated in the body, the methylation patterns can also act as a blueprint for where the DNA comes from."It's very much like a message in a bottle," says Jasmine Zhou, a professor of pathology at the University of California, Los Angeles, and co-lead author of the study. "This cell-free DNA floating in the blood can tell us the secrets of each cell or organ that it's from."So far, only one FDA-approved test is available that uses this type of free-floating tumor DNA, and it relies on fecal samples, not blood. Other non-invasive screening methods test for intact tumor cells that are loose in the bloodstream. But finding intact tumor cells in blood often only gives an indication of cancer prognosis and is like searching for the proverbial needle in a haystack.Advancements in gene-sequencing machines, which rapidly decode millions of fragments of DNA, have driven the hunt for new screening tests over the last five years. By detecting tumor DNA, instead of intact cells, researchers around the world have been trying to spot signs of cancer and monitor changes in tumors in real time.
Zhou and her colleagues added the power of machine learning to a blood test. They drew on data available in the Cancer Genome Atlas database to teach the CancerLocator software to recognize which patterns were normal, and which ones were tumor signatures. This allowed the researchers to accurately pinpoint cancer even when the amount of DNA sequenced was very low.In blood tests from 29 liver cancer patients, 12 lung cancer patients and five breast cancer patients, the program was able to reduce errors in detecting cancer from roughly 60 percent to 26.5 percent.In some cancers, this computer algorithm could be particularly advantageous, Zhou says. For example, if there is only a small amount of tumor DNA in the blood — as is usually the case in early stage cancers — or if a tumor is close to a major blood vessel.In addition to offering clues about the presence and location of a tumor, the program could also be used to inform personalized treatment options, monitor the effectiveness of a drug and give an early warning about possible recurrence.As researchers learn more about the tumor signatures, their genetic data can also be plugged back into CancerLocator's machine learning program to further improve the test's diagnostic accuracy.There's definitely an incentive to improve early cancer diagnosis. It's the most important factor in increasing a patient's odds of survival. And many pharmaceutical companies have already started offering drugs that come with their own required DNA tests. Moreover, if CancerLocator proves effective at accurately diagnosing a wider range of cancers, in a larger group of people, it could be used for regular cancer screening in healthy people too, Zhou says."That's the long-term goal—to develop this into a screening test because the potential benefits to the public are huge."