Skip to main content
Premium Trial:

Request an Annual Quote

Grail Unveils ctDNA Atlas Study to Support Development of Cancer Early Detection Test


NEW YORK (GenomeWeb) – Liquid biopsy firm Grail said last week that it is moving forward with an ambitious plan to comprehensively characterize circulating DNA in at least 10,000 individuals — the first details it has released on how it will develop a planned pan-cancer early detection test.

The goal of the study is to create a circulating cell-free genome atlas (CCGA), which describes and compares the heterogeneity, or variability of ctDNA in both healthy people and patients with cancer. This will then support Grail's test development goals, allowing it to precisely tune its assay to identify patterns that indicate a developing tumor in an otherwise healthy person.

Launched this January by sequencing titan Illumina, Grail did not share much detail at first about its plans or technology, save that it was taking the long view on cfDNA-based cancer early detection, expecting to spend years on research to meticulously validate its test.

This view contrasts with that of some other companies, which have already launched blood-based cancer screening tests.

For instance, Pathway Genomics last year introduced a test called CancerIntercept Detect for the screening of asymptomatic but at-risk individuals.

Although the US Food and Drug Administration sent Pathway a letter raising concern over a lack published evidence that CancerIntercept or any similar test had been adequately validated, the company is still marketing the assay on its website.

More recently, testing firm CellMax life has begun to offer a circulating tumor cell-based test for colorectal cancer screening in Asia, which it plans to also launch in the US.

Others have tracked closer to Grail in adopting a more methodical, longer-term strategy. Guardant Health, for example, announced its own multi-site study this June.

In that effort, called called Project LUNAR (liquid biopsy using next-generation sequencing to assay high-risk patients), Guardant is partnering with its own slate of academic institutions to recruit patients from high-risk populations or with early-stage disease in whom it will validate the ability of its own targeted sequencing test to detect nascent tumors.

Unlike Grail, which is working now on building its atlas, Guardant has already sequenced ctDNA in tens of thousands of cancer patients in the context of its Guardant360 test, which is intended for analyzing the makeup of cancer in patients who have already been diagnosed.

Because of that, it believes it is ready now to begin collecting validity data on the sensitivity of its technology to detect cancer at its earliest stages.

Grail is starting somewhat more from scratch, but, with a much broader and deeper sequencing methodology it has set itself up to potentially push the limits of detection even further than Guardant to detect shreds of DNA that might mark the invisible first inklings of tumor development.

To build its CCGA, Grail said that it is collaborating with a number of academic centers to recruit a total of 10,000 subjects whose blood samples it will analyze using a comprehensive targeted panel and an ultra-deep sequencing strategy. The company plans to follow patients for up to five years, tracking changes in their circulating DNA patterns and levels.

Within the cohort, 7,000 subjects will be cancer patients, and another 3,000 will be healthy controls.

Though he did not provide details of exactly what areas of the genome are included, Grail's CEO Jeff Huber told GenomeWeb that the company will be using a sequencing assay that covers hundreds of genes — an order of magnitude larger than what other companies in the space have described so far.

Huber also said that the firm's testing strategy involves sequencing an order of magnitude deeper than it is aware of any other company doing.

"If you think about the challenges of detecting cancer at early stage, there are two main dimensions," Huber said. "One, is that cancer is spectacularly heterogeneous … Given that heterogeneity and the fact that we want to detect cancer at its earliest stages, we have to look very broadly across the genome to figure out where mutations are happening and be able to detect them," he said.

"The other dimension is depth … The goal is to be able to detect in a full tube of blood literally a handful of molecules that are indicative of cancer at its earliest stages …  so we are sequencing an order of magnitude deeper than anyone else we are aware of as well."

"If you multiply that breadth and that depth, we are generating on the order of a terabyte of data per test subject," Huber added.

So far, Grail has named a handful of collaborators, including the Guardian Research Network, Mayo Clinic, and Hartford HealthCare Cancer Institute, but it said that CCGA study will eventually include up to four dozen clinical trial sites.

With the first 10,000-subject phase of the study, the company will be focused on directly comparing the heterogeneity or diversity of circulating cfDNA patterns in healthy people to those in people who have cancer of various stages and types.

The initial phase won't include testing of the assay in at-risk but otherwise healthy people like Guardant is doing, Huber said. That will come later, when the company moves on to validate the ability of its panel to identify the earliest signals of developing cancer.

"One of the concerns whenever you have a diagnostic test is what are your false positive and false negative rates. So focusing on a large healthy control population, we can make sure we have a very low false positive and false negative rate."

Huber reiterated what Grail has maintained from it's founding, which is that defining clinical utility for the test will eventually require validation in not just 10,000 but likely hundreds of thousands of individuals.

As the liquid biopsy space has crowded, at least in the area of analyzing patients who already have cancer, many companies now also distinguish themselves based on proprietary or "secret sauce" methods for error correction that they claim allow detection of molecules at miniscule frequencies among background normal DNA.

According to Huber, Grail has its own recipes in that area as well, which he said the company believes are industry-leading due to their development by scientists working on Illumina's core sequencing technology itself.

However, he said, "the thing that motivates us is saving lives and I don't think there is any monopoly on that, so we don't view others as competitors in that respect."

Grail does believe it is tackling challenges that other companies are not, though, especially in its goals for sequencing hundreds of thousands of subjects and its work behind the scenes to develop the informatics and data infrastructure to support the eventual fruit of that.

"People talk about big data in healthcare, but its kind of a misnomer," he said. In a lot of cases the quantity and sample sizes are actually relatively small, compared to other industries.

"For us … if we successfully create a product that is deeply integrated into the medical system and is being used for screening in the US adult population, that's about 100 million people a year," Huber said. "At a terabyte each, that quickly makes you the first yottabyte-scale [healthcare] application."

To that end, Huber said that Grail is laying groundwork now for the infrastructure, machine learning, and other capabilities necessary to solve that. "We haven’t seen anyone else doing what we are doing in that respect," he said.

A lingering question for Grail as it moves forward will be whether a test that relies on broad and deep sequencing can be offered at a cost appropriate for screening large numbers of asymptomatic people.

Huber said though that the company is confident that if costs continue to shrink along the lines that they have been — both on the sequencing technology side, and the compute/data storage side — it will be in a good position.

"We couldn’t launch the test today … but we are confident in a launch time window on the order of four to five years, that costs will be able to get to points that work well with healthcare economics," he said.