The DNA testing company Color Genomics says that early next year it will offer customers a new type of gene analysis that could tell them if they are at risk of a heart attack.
Here’s the problem, though: it only works in white people.
The new type of DNA analysis, called a polygenic risk score, can predict people’s risk for a variety of serious diseases, including cancer, diabetes, and clogged arteries. It’s a big enough deal to be on our list of 10 Breakthrough Technologies this year.
But because the algorithms were developed by using large databases of health and DNA data mostly from people with of recent European ancestry, the tests aren’t as accurate in other groups, like black Americans, Hispanics, and Asians.
For companies like Color, based near San Francisco, that poses a dilemma.
“To be only able to offer a product to one part of population is inequitable and should be unacceptable to the field,” says Alicia Zhou, head of research at the company. On the other hand, the tests do work in some people; they could actually save lives now. "Holding back data from individuals that could benefit," would also be wrong, she says.
To help solve the problem, the company says, it has made genetic data available on 50,000 people, including minorities. And it hopes other scientists will use the data to speed the work of developing risk scores for people of all backgrounds. “We can’t wait for five years,” says Zhou.
Traditional genetic tests look deeply at single genes for mutations that cause disease. The new risk scores work differently. Instead of inspecting one gene, they consider a large number of measures from across a person’s genome—usually about 500,000 locations—to create a composite score reflecting more risk or less, depending on the overall result.
The problem is that biobanks employed to prepare the scores largely contain DNA from people of European ancestry. As of 2016, about 80% of the 35 million participants in one set of big science projects were of European descent. Hispanics were just 1%. (See “DNA databases are too white. This man aims to fix that.”)
That means the DNA scores—and the risks they imply—aren’t nearly as accurate when applied to other ethnicities, whose ancestral genetic patterns are different. Those underlying differences decrease the accuracy of the risk models, sometimes so much that they might not be safe to use.
The result, geneticists fear, is a race-specific medical advance.
Some scientists are racing to fix this problem. For instance, Alicia Martin of the Broad Institute in Cambridge, Massachusetts, and 22 colleagues presented a statistical scheme for adjusting risk calculations to increase their accuracy in different populations.
“The fact that [risk scores] offer far greater predictive value in individuals of recent European ancestry than others is perhaps the primary ethical and scientific challenge preventing their clinical implementation,” Martin and her colleagues write.
Even so, race-restricted polygenic risk tests are already for sale. In September 2017, for example, Myriad Genetics of Salt Lake City added a new “riskScore” to its well-known breast cancer test. In the past, Myriad only analyzed infamous genes like BRCA2 to look for mutations that are known to lead to cancer. The new test, it believes, can identify at-risk women who don’t have an inherited BRCA mutation. It employs a scoring system to assess variations at 86 locations in the genome at once.
Yet because Myriad’s research involved people of European background, it had to restrict the new test only to them. Similarly, Ambry Genetics, which offers an “individualized” assessment of a man’s risk for prostate cancer, says only people of “Northern European ancestry” are eligible.
Ron Rogers, a spokesman for Myriad, says the company is working to expand the test so it serves "multiple ethnicities.” He said by early December the company expects to announce at a meeting in San Antonio, Texas, that it has proved the test works in Hispanics as well. “The science continues, and we are working on it as quickly as we can,” says Rogers.
The validity of the polygenic risk scores is still widely debated, and it’s not clear how much they will help doctors or patients. In the case of coronary artery disease, though, the new approach is able to find about 10 times more people at high risk than conventional assessments do.
Those individuals could be given cholesterol-lowering drugs, says Sekar Kathiresan, a cardiologist at Massachusetts General Hospital and the Broad Institute, who developed the heart score, a version of which Color plans to implement.
Since the heart test could save lives, Kathiresan is eager to make the information widely available. In addition to helping Color prepare its test, he says he plans to let people known their risk for free through a web portal (consumers would upload their 23andMe profiles to receive a result).
Kathiresan agrees that the score “does not predict risk as well in blacks as whites.” To improve it, he says, additional large heart studies need to be carried out “in each ethnic group” to generate ethnic-specific scores.
Color is taking a different approach. Currently, the risk scores are calculated by measuring DNA on so-called microarray chips, which cheaply measure about 500,000 to 1 million individual letters in a person’s genome. However, these chips are also designed with a bias toward the genomes of Europeans, says Zhou.
In an announcement today, Color says it is has been able to gather the same information using high-speed DNA sequencers, the type of machinery it already employs in its lab. The cost is less than $40, according to Othman Laraki, the company’s founder. Laraki says the innovation will allow Color to offer a heart attack risk test early next year and may eventually help resolve racial disparities, too.
“The limitation of not being able to apply it to different populations is a thorn in side of the field,” he says.
Be the leader your company needs. Implement ethical AI.
Join us at EmTech Digital 2019.