The posters in the clinic at Gibbs Cancer Center, a regional hospital in Spartanburg, South Carolina, say “Help Stop Cancer.” Volunteers who agree to donate blood will receive a $25 gift card and their donation will help “determine if a new blood test can be used to detect cancer earlier.”
The posters are paid for by Grail Inc., a Silicon Valley startup that last year began laying out a bet that the “cure” for most cancers isn’t a new drug, but tests that catch tumors at an early stage when they can still be successfully treated. Grail thinks it can do that using “high-intensity” DNA sequencing to probe blood samples for genetic material shed by hidden tumors.
Grail has raised $1.1 billion, a sum that puts it among the top three most heavily funded private biotech companies, according to EvaluatePharma. The company will need that money and more to win a technological race that is as significant, and as costly, as those toward driverless vehicles or new forms of artificial intelligence. Investors include gene-sequencing giant Illumina, Johnson & Johnson, and Amazon founder Jeff Bezos.
Huge and expensive studies will be needed to discover the fingerprints of cancer and then prove a “screening” test really helps. Grail last year began seeking blood from 7,000 cancer patients at Gibbs and other community hospitals, and in April began exploring whether it can improve on mammograms, an x-ray of the breast that women are advised to get annually starting at age 45, by finding breast cancers earlier, and finding more of them.
To do so, Grail is now seeking 120,000 women to provide blood when they undergo mammograms at centers including the Mayo Clinic. About 650 of those women, cancer statistics forecast, will go on to develop breast cancer within a year of entering the study. Grail will then analyze their banked blood to determine whether a DNA test could have correctly “predicted” their cancer from a blood draw.
The study involves so many people that just the $25 payments to patients will run Grail $3 million. “The hospitals want to know unequivocally that we are funded,” says Grail’s chief business officer, Ken Drazan, who has led fundraising for the company. “They don’t want to take the risk that we won’t be able to follow through.”
Grail is headed by Jeff Huber, a former Google software executive whose wife, Laura, died of colorectal cancer shortly before he joined the company as CEO. Laura’s death, one of approximately eight million from cancer that year around the world, provides moral direction for Grail, and her $2 million insurance bill explains its economic rationale.
Laura’s life could have been saved, Huber says, with the type of test Grail is developing, followed by an inexpensive surgery. “Things don’t happen for a reason,” Huber told a commencement crowd at the University of Illinois in May. “But they do often happen because nobody has yet found a better way.”
Other biotech companies, like Guardant Health and Personal Genome Diagnostics, are already selling DNA tests for people who are fighting cancer, or are recovering and want to catch a relapse. But Grail’s goal of screening asymptomatic people is more challenging because a small cancer is hard to detect, and there could be little room for wrong results, or false positives, which could cause people to panic.
Aiming straight for an early-detection test is like “going to the top of Everest without passing through base camp,” says Helmy Eltoukhy, head of Guardant Health, a competitor that has raised nearly $600 million and sells blood tests to people with cancer so they can pick the right drug. But he agrees screening millions at checkups is a real Holy Grail, and says Guardant is moving in that direction too. “It’s the single biggest diagnostic opportunity in recent history.”
The argument in favor of screening for cancer seems obvious: two-thirds of cancers caught early are curable, often with surgery. Of those caught late, 80 percent are fatal. But screening tests have notorious downsides. They can set off false alarms or cause patients to pursue unnecessary treatment. “The big problem with this technology is not going to be finding cancer, it will be finding the cancers that should be treated,” says Laura Hercher, a genetic counselor who teaches at Sarah Lawrence College.
Huber says tests developed by Grail will save money and will do more than tell doctors whether a cancer is present. In addition, he told Technology Review, the company’s tests will have to tell doctors where the cancer is in the body, how aggressive it is, and, if already advanced, what treatment to seek.
Until recently, Grail seemed unlikely to meet its objective of having its first test on the market by 2019. That changed last week when the company announced it would combine its operations with those of Cirina, a screening startup created by Hong Kong molecular anatomist Dennis Lo, and one of the scientific godfathers of the blood test idea (see “10 Breakthrough Technologies: Liquid Biopsy”).
Though tiny by comparison—it had raised $12 million in October—Cirina controls key patents filed by Lo, who with colleagues at the University of Hong Kong has already spent 17 years determining that a blood test can detect a cancer called nasopharyngeal carcinoma, which is common among men in Hong Kong and Southern China and is caused by the Epstein-Barr virus.
A Grail spokesperson says Lo’s team at the University of Hong Kong is now submitting a description of their years-long study of 20,000 men to a journal. They will report that blood tests allowed the nasal cancer to be found earlier and cured more often, providing what may be the first complete proof that liquid-biopsy screening tests can save lives. Cirina had been planning to introduce a commercial screening test in Hong Kong this year.
At Grail’s 70,000-square-foot headquarters in Menlo Park, California, half the space is used for labs and a platoon of ultrafast DNA sequencing machines. These devices—each cost $1 million—are what allow Grail to deeply analyze a blood sample for infinitesimal fragments of DNA shed by tumor cells.
These fragments, if spotted early enough, are what can flag the presence of a cancer before a person feels any symptoms.
Scientists have known there is free-floating DNA in the blood since the 1960s. But speedy “next-generation” sequencers only recently made it practical to analyze. The technology’s first success came in 2011. That’s when companies launched new “noninvasive” prenatal tests that scan a pregnant woman’s blood for bits of DNA from her fetus. Quick and accurate, the tests quickly became a standard choice for parents to learn if their unborn child has Down syndrome or other serious genetic errors.
But some tests began showing odd results—in fact, they had cancer, and the baby test was picking it up. “That lit the light bulb,” says Huber, Google’s representative on Illumina’s board of directors at the time. “We were seeing the signal. What would it take to do it right?”
At Grail, blood samples undergo what it calls high-intensity sequencing. The data then flow into computers that parse, analyze, and store the information—a software “pipeline” whose complexity makes Grail as much a software enterprise as a biotech.
In fact, Grail says it collects 1,000 gigabytes of raw data from each volunteer’s blood, equal to about 500 hours of movies. That, along with information from people’s medical records, is fed into a “classifier,” the software that will be searching for patterns. Huber says if such testing ever became routine, Grail would “quickly become the first zettabyte operation,” or the biggest big data company in the world.
Recommended for You
But first Grail has to discover which patterns flag a hidden cancer. According to Grail’s scientists, the DNA data are stuffed with clues, but exactly which features matter most is still to be determined. It could be mutations in the DNA, the size of the fragments, chemical modifications found on the genetic strands, or a combination.
“We have not landed on a single strategy that we are taking forward,” says Anne-Renee Hartman, head of clinical development at Grail. “The work that’s been done and published is on people that have cancer. We have to employ every strategy.”
Going for broke
Silicon Valley’s approach to biology leaves some observers skeptical. “The lets-measure-the-hell-out-of-it-with-big-data approach ignores how complex biology is,” cautioned one doctor, who asked for anonymity because he works for a Grail competitor. “In biology, we don’t even know what we’re supposed to be measuring.”
One problem is that signals of cancer may not always be present. Luis Diaz, a cancer scientist at Memorial Sloan Kettering in New York, thinks that in 35 to 40 percent of cancer types there’s essentially nothing to find early-on because they don’t shed enough DNA into the blood. In other cases, the amount of cancer DNA might be at the limits of what’s detectable. “It’s the weak-signal problem,” says Guardant’s Eltoukhy. “That’s why the studies are so big.”
Huber also says Grail’s high-intensity sequencing tactic is not yet cheap enough to deploy as a doctor’s office test. To find rare mutated copies of a gene expelled by a cancer cell in an ocean of normal copies, each is sequenced 75,000 times. Right now, you could buy a used car for what Grail’s test costs to perform on a single blood sample.
And the tests need to be incredibly accurate. About 1 in 200 Americans will receive a cancer diagnosis in 2017. If millions of healthy people were screened, even a low rate of false positives could create an epidemic of anxiety and cancer scares.
Even a cancer detector that works perfectly could cause social upset. What if it’s expensive and only for rich people? What happens to people who don’t have much access to cancer care? What’s more, in the real world, finding cancer early sometimes has doubtful benefits. The value of the recommended screening test for prostate cancer, in use since 1990, remains widely debated. No one is sure if finding these cancers actually saves lives overall or if it wastes resources and causes harm to men who get treatment for cancers that wouldn’t harm them.
“If [Grail reports] every cancer, we will go broke. That is clear as day,” says Hercher. “We can’t infinitely add to the most expensive health care in the world.”
Huber says Grail’s goal is unchanged: a mass-market screening test cheap enough for wide use. To have the big effect on cancer deaths that Grail is seeking, he adds, it “has to happen at a population scale, and we have to make it affordable for the system to do that.”
Grail is already convening insurers, economists, and regulators to discuss how society can pay for the tests and incorporate them into preventive medicine programs. Huber wouldn’t cite a price tag for the technology, but says, “We are not going to offer a $169 test on day one.” More likely, tests could cost $1,000 at first. If so, screening every American just once would cost $300 billion—not counting the cost of follow-up.
Will anyone pay? Huber says his wife’s costly treatment with drugs that didn’t save her shows why they would. “It’s a huge amount spent by the system where the chance of a positive outcome is vanishingly low,” he explains. “Whatever price you want to pick for a Grail test, if it had caught her cancer at stage two, she could have been cured with a blunt instrument, a $10,000 or $20,000 surgery. We believe there is a lot of room in the system to make this work.”
AI is here. Will you lead or follow?
Join us at EmTech Digital 2019.