What if instead of invasive cancer tests, scientists could run a blood sample through a computer program and not only detect whether cancer is present or not, but pinpoint where in the body it’s located? This technology, harnessed by a program called CancerLocator, could potentially be ready in a year.
In a recent study published in Genome Biology, researchers from the University of California at Los Angeles (UCLA) developed a computer program that identifies specific epigenetic patterns, or a combination of chemical marks on DNA, that are associated with certain cancers. The tool functions by measuring circulating tumor DNA in blood and comparing epigenetic marks to a database.
“Non-invasive diagnosis of cancer is important, as it allows the early diagnosis of cancer, and the earlier the cancer is caught, the higher chance a patient has of beating the disease. We have developed a computer-driven test that can detect cancer, and also identify the type of cancer, from a single blood sample,” said the co-lead author from UCLA, Professor Jasmine Zhou.
“The technology is in its infancy and requires further validation, but the potential benefits to patients are huge,” she added.
DNA from tumor cells can be found in the bloodstream during the earliest stages of cancer and can be a unique target for detecting the disease just as it starts to form. The computer program works by searching for particular epigenetic patterns in cancer DNA, specifically DNA methylation marks, which flow freely in the patients’ blood. These epigenetic signatures, found in what is called circulating cell-free DNA (cfDNA), were cross referenced to a database of patterns from different types of cancer which had been collected by the researchers.
“We built a database of epigenetic markers, specifically methylation patterns, which are common across many types of cancer and also specific to cancers originating from specific tissue, such as the lung or liver,” Zhou explained. “We also compiled the same ‘molecular footprint’ for non-cancerous samples so we had a baseline footprint to compare the cancer samples against. These markers can be used to deconvolute the DNA found freely in the blood into tumor DNA and non-tumor DNA.”
The novel computer program was tested alongside two other methods, using a variety of blood samples including those from patients with breast cancer, lung cancer, and liver cancer. The other methods, Random Forest and Support Vector Machine, had a higher error rate of 0.646 and 0.604, respectively, compared to the new program which had a low error rate of 0.265. This means that it was rare for CancerLocator to indicate that a patient had cancer when they actually did not.
“I hope it [a diagnostic test] will be available within a year. It depends on training data, testing and machine learning,” Zhou told The Independent. “With enlarged training and testing data we are confident to achieve much higher performance.”
In this study, 25 out of 29 people with liver cancer and 5 out of 12 people who had lung cancer had early stage cancers. In 80% of the cases, the program was able to detect the disease even though tumor DNA levels are typically lower in the earlier stages.
“In general, the higher the fraction of tumor DNAs in blood, the more accurate the program was at producing a diagnostic result,” Zhou explained. “Therefore, tumors in well-circulated organs, such as the liver or lungs are easier to diagnose early using this approach, than in less-circulated organs such as the breast.”
Although only 3 cancer types were tested with this diagnostic computer program due to limited blood samples, the epigenetic tool remains a viable option for noninvasive cancer detection.
Source: Kang, S. et al. (2017). CancerLocator: non-invasive cancer diagnosis and tissue-of-origin prediction using methylation profiles of cell-free DNA. Genome Biology, 18:53.
Reference: Computer program developed to diagnose and locate cancer from a blood sample. Science Daily. 24 Mar 2017. Web.