Bay Area GRAIL's Billion Gamble on the Perfect Cancer Test

Bay Area GRAIL's Billion Gamble on the Perfect Cancer Test June 6, 2017
By Mark Terry, BioSpace.com Breaking News Staff

Wanted: blood samples from 120,000 women who are undergoing mammograms.

That’s one of the immediate goals of biotech startup, Grail Bio, which was launched in January 2016 by San Diego-based Illumina , a leader in DNA sequencing technology. Grail’s goal is to develop an early screening test to identify cancer. And it’s currently raised $1.1 billion to do it. Investors include Illumina, Johnson & Johnson , and Jeff Bezos, the founder of Amazon .

Last year, the company began acquiring blood samples from 7,000 cancer patients at Gibbs Cancer Center in Spartanburg, SC and other community hospitals. And this April it stated that it wanted to create a test to compete with or replace mammograms. Which is where the 120,000 blood samples come from—a variety of hospitals and medical centers, including the Mayo Clinic. They are currently paying patients $25 for each blood sample, which will add up to around $3 million.

“The hospitals want to know unequivocally that we are funded,” Grail’s chief business officer, Ken Drazan, told the MIT Technology Review. “They don’t want to take the risk that we won’t be able to follow through.”

In many ways, Grail’s approach is quintessential Silicon Valley—if we can just acquire enough data and analyze it, a solution will present itself. And Grail isn’t the only company using this approach, although it may be the best funded. Others include Guardant Health and Personal Genome Diagnostics, and it seems likely that Google ’s Verily, its life science company, is using some variations of this approach for some of its research projects.

The rationale for an early screening test is simple—about two-thirds of cancers are curable if they’re caught early. But when they’re caught late, about 80 percent are fatal.

The problem though, is complexity. Another word for it, and it’s one that both biologists and data scientists are familiar with, is “noise.” A blood sample has a lot of stuff in it—blood cells, immune cells, plasma, proteins, enzymes, etc. And much of it—most of it, really—is noise that has nothing to do with cancer.

As the MIT Technology Review notes, scientists have known about free-floating DNA in the blood since the 1960s, but it’s only recently with next-generation sequencing technology that analyzing it would be practical. Grail says it’s currently collecting 1,000 gigabytes of raw data from each blood sample, “equal to about 500 hours of movies.” They also collect information from the patients’ medical records. All of this data is then entered into a “classifier,” or software that focuses on pattern recognition.

What they’re looking for are patterns that signal a hidden cancer. And in fact, the information—at least some of it—is likely there … buried in the noise.

“We have not landed on a single strategy that we are taking forward,” Anne-Renee Hartman, head of Grail’s clinical development, told MIT Technology Review. “The work that’s been done and published is on people that have cancer. We have to employ every strategy.”

On June 3, Grail announced results from a study that evaluated its high-intensity sequencing approach on blood samples. Blood and tumor samples from 124 patients with various advanced cancers were sequenced with a 508-gene panel. It yielded about 100 times more sequencing data than earlier approaches. In 89 percent of the patients, the company correlated at least one of the mutations seen in the tumor tissue as also observed in the blood. The company stated, “When evaluating all genetic variations, including those present at high levels in tumor tissue as well as those at low levels, 627 of 864 mutations (73 percent) detected in tumor tissue were also detected in patients’ blood.”

There are other challenges. One physician, unidentified because he works for a Grail competitor, told MIT Technology Review, “The lets-measure-the-hell-out-of-it-with-big-data approach ignores how complex biology is. In biology, we don’t even know what we’re supposed to be measuring.”

A sentiment echoed by Luis Diaz, a cancer researcher at Memorial Sloan Kettering in New York, who notes that about 35 to 40 percent of cancers types don’t shed DNA into the blood, so aren’t likely to be identifiable. “It’s a weak-signal problem,” Guardant’s chief executive officer, Helmy Elktoukhy, told MIT Technology Review. “That’s why the studies are so big.”

The technology, if it works, also presents other potential issues. They include whether and how to treat every early cancer identified, and cost. Laura Hercher, a genetic counselor who teaches at Sarah Lawrence College, told the Review, “If [Grail reports] every cancer, we will go broke. That is clear as day. We can’t infinitely add to the most expensive health care in the world.”

Time will tell. And Illumina was able to bring full-body gene sequencing under $1,000. But the company’s goal of having a test on the market by 2019 is daunting.

Back to news