Nick Papadopoulos tracks down tumors for a living. Not with X-rays or CT scans, but with DNA. The oncologist and director of translational genetics at the Johns Hopkins Kimmel Cancer Center has spent decades uncovering the unique sets of mutations that define cancers—the kind of genetic signals that not only drive tumor formation and metastasis, but distinguish one cancer from another. And now, he’s working to develop a test that could sniff out those signals before a patient starts to get sick.
It’s the kind of test that Papadopoulos thinks could have saved his uncle’s life, had it been around a few years ago. “He had no symptoms until a cough showed up,” he says. But when it didn’t go away he went in for an X-ray, and there on the radiograph were the lesions. Dozens of them, filling his entire chest cavity. The doctors sequenced the tumors, and got him signed up for a clinical trial for a new, targeted drug. It worked for a few of them, shrinking them back to almost nothing. But the rest developed resistance.
“He was supposed to only live two months, and the drugs prolonged his life by a year. But that year wasn’t good.” says Papadopoulos. “I think it’s time to start thinking more about detecting cancers early and less about treating them when they are late.”
On Thursday, Papadopoulos’ research group at Hopkins revealed a novel blood test based on the combined analysis of DNA and proteins that correctly detected eight kinds of the most common cancers with a range of accuracies—from 98 percent for ovarian cancers to less than 40 percent for breast cancers. Published in Science, the test is just one among many so-called “liquid biopsies” in development; noninvasive tests that classify cancers by identifying the tiny bits of DNA that tumors shed into the bloodstream.
Most published studies, including this one, focus on measuring and monitoring advanced tumor stages. A few liquid biopsies have even been approved to help match tumors to targeted drugs. But the dream is to develop a simple blood test to actually diagnose solid tumors in healthy-looking people. The scarcity of circulating cancer biomarkers (both in quality and quantity; tumor DNA makes up less than 0.1 percent of blood) has held those aspirations back for decades. But now, sensitive assays and computational platforms are driving the discovery of biomarkers and better ways to measure them, luring a pack of well-financed startups into the field.
In 2016, for example, the world’s largest sequencing company, San Diego-based Illumina, spun out a new company called Grail. Its mission is described as “detecting cancer early, when it can be cured.” This ambitious aim is supported by $1.2 billion of venture capital Grail raised last year, which it intends to put toward financing massive, population-based clinical studies and optimizing its sensitive sequencing technologies.
Grail has yet to publish any actual data (its website does advertise a commentary published in Cell last year). And neither has its chief rival in the Valley, a machine learning startup called Freenome. That three-year old company snagged a $65 million Series A last March, led by Andreessen Horowitz. Freenome isn’t limiting itself to the genetic breadcrumbs left by tumor cells—it looks to capture other disease signatures in the blood, like how the immune system changes in response to tumor microenvironments.
Of course, Freenome has offered scant details on how exactly that kind of test would work. “You show your cards at the end, not while you’re playing poker,” says Andreessen partner Vijay Pande, who heads the investment firm’s biofunds. “Publications indicate that you’re not interested in building a company.” That said, he does expect Freenome to publish in a peer-reviewed journal ahead of its first foray into the market.
When that could be, though, is anyone’s guess. To evaluate any of these blood screens, thousands of patients will have to get tested—and then researchers will have to wait for some of them to actually get cancer. That’s the only way to determine not only their predictive power, but also whether they lead to improved patient outcomes. The noninvasive screening tests available today—mammography for breast cancer, a protein-measuring test for prostate cancer—are rife with their own issues. Incorrect diagnoses waste time and money on treatments and burden patients with unnecessary anxiety.
More on Liquid Biopsies
Liquid biopsy is likely to be beset by the same kinds of controversy, says Geoff Oxnard, a thoracic oncologist at the Dana-Farber Cancer Institute and a professor at Harvard Medical School. He routinely uses a single-gene liquid biopsy developed at Dana Farber to figure out which drugs represent the best options for his lung cancer patients. But will early detection versions one day be part of routine doctor’s visits? “No. I think these tests will help us better understand the risks for patients who already have a history of cancer in their family or who’ve already had something show up on a scan,” he says. “But I don’t think we have the kind of data we need to support liquid biopsy as a panacea for diagnosing cancer. At the end of the day, it’s still just a shortcut.”
Still, Oxnard pointed out that Papadopoulos’s test represents an important step forward. One, it starts to identify where a tumor might be located. That’s been a big limitation of liquid biopsies; OK, you’ve found cancer, but what do you do next? Where do you look for the tumor? Most mutations don’t tell you anything about location. But by layering in measurements for 31 additional proteins to their machine learning model, the Hopkins team was able, on the first try, to correctly identify the tissue of origin around 80 percent of the time colorectal cancers, pancreatic, and ovarian cancers.
The other advance is cost. Papadopoulos estimates the test could be commercialized for around $500, and cancer-spotting approaches that rely on ultra-deep sequencing could stretch costs for existing screening tests, which only look for a single gene. “This is great for the field and provides promise that these analyses will become a reality in the clinic,” says Victor Velculescu, an oncologist and colleague of Papadopoulos’ at Johns Hopkins, who has also developed liquid biopsy technologies, though he was not involved in the Science study.
The two have developed a sort of friendly turf war as they’ve turned Baltimore into its own little liquid biopsy hub. Both researchers have recently spun off diagnostics companies to further develop their own early detection technology platforms. Earlier this month, Velculescu’s venture, Personal Genome Diagnostics, hauled in a $75 million Series B led by pharma giant Bristol-Myers Squibb. That brings its total financing to $99 million, putting it on par with some of its better-known counterparts in the Valley, adding some bicoastal intrigue to the race to the market. Whatever the outcome, it’s patients who will ultimately be the winners.
“If it can even catch 50 percent of cancers that right now we have no way of screening for, that’s still 50 percent of patients who can now be treated in Stage 1, when they still have a chance,” says Papadopoulos. “It doesn’t have to be perfect to still save a lot of lives.”
