Taking the Uncertainty Out of Genetic Screening for Cancer Risk

In 2013, the U.S. actress Angelina Jolie learned that she had mutations in a gene called BRCA1, which conferred very high risks of breast and ovarian cancer. To obviate that risk, she opted to have a double mastectomy; in 2015, she had her ovaries removed, too.

Many women face similar choices on the basis of BRCA1 tests, which search the gene for mutations that are known to increase the risk of breast cancer. Geneticists have documented many such mutations, as well as those that are more benign. But BRCA1 is a big gene, and there’s a lot about it that we don’t know.

This means that a test might return with a “variant of unknown significance” or a VUS—mutations that, simply put, we know squat about. They could ramp up the odds of cancer, by some unknown degree. Or they could do nothing. For BRCA1, there are at least 350 VUS in total. And around 2 percent of women who go for the most widely used BRCA1 test, offered by Myriad Genetics, will see at least one of these unknown mutations. Their presence on a set of test results is a big lingering question mark, an admission of ignorance, a disquieting clinical shrug.

“Women who carry a VUS seek to make informed decisions about their bodies and their health,” says Andrea Downing, an activist and BRCA1-mutation carrier. “It’s very hard to do this without clarity about whether a VUS is harmful or benign.”  

To get that clarity, you could do painstaking experiments to work out what each of these variants does, but “clearly, no one’s willing to do that because there are still 350 of them around,” says Lea Starita from the University of Washington. Alternatively, geneticists could rely on computer programs that predict whether a variant is likely to be harmful or benign, which would be much faster but also terribly unreliable.

Starita presented the approach at the American Society for Human Genetics 2015 conference. Here, the halls are full of scientists who start with a disease, look for a genetic variant that’s associated with it, and then characterize it. It is worthy but laborious work, which generates a seemingly endless array of thematically similar talks and posters. Starita’s team has flipped this approach around: They’re starting with genes of great interest and working out what happens if they mutate it in every possible way. Let’s characterize every possible variant, they’re saying, including ones that have never been seen before.

Or in Rumsfeldian terms, they’re going after the unknown unknowns.

Starita focused on a critically important segment of BRCA1 called the RING domain, which is especially rich in cancer-predisposing mutations. The RING domain codes for a string of 120 amino acids, and each of these could be substituted for 19 alternatives. Starita created all of these substitutions, producing a library of 2,040 mutants. Together with colleague Stan Fields, she simultaneously put the whole library of mutants through a the same biochemical tests, and scored them according to how well they perform their typical jobs.

The team then developed a way of converting the scores in these tests into predictions of risk, which say whether a particular mutation is likely to be harmful, benign, or somewhere in between. And their predictions not only include several VUS, but also 1,287 variants that have never been seen before. “In these genes of strong interest, we’re justified in making a concerted effort to test every possible mutation, even if it hasn’t been seen in a patient yet,” says Jay Shendure, who co-led the study. “Because some day it will be. Then we’ll have an interpretation.”

The results aren’t ready for clinical use yet, because the team still needs to confirm the accuracy of their predictions. That will take time, and some basic epidemiology: Simply put, are women who have variants identified as risky actually more likely to develop breast cancer? They also need to extend their technique across the entire BRCA1 gene, in all its enormity. Meanwhile, they are already applying their technique to other important genes, including more cancer-related ones, and others that affect a person’s reaction to medical drugs.

It’s an ambitious approach, of a kind that will be more important as genetic testing becomes more commonplace. As Starita notes, it’s a big business. There are more and more companies, testing for more and more genes, and finding increasingly many variants of unknown significance. Some tests, which look at panels of dozens of genes, find at least one VUS around 40 percent of the time.

Many companies are also hoarding data on important variants. Myriad Genetics, which had long been embroiled in a legal battle over its attempt to patent the BRCA1 and BRCA2 genes, is also sitting on a proprietary database of BRCA1 variants that it does not share with researchers.* “This is a huge part of the VUS problem,” says Downing. “Our community’s genetic data shouldn’t be a trade secret—it should be published to ClinVar,” an open database on disease mutations.

“Myriad supports the goal to improve the safety of public databases by continuing to contribute variant classifications and techniques through peer reviewed publications,” the company said, in a statement to The Atlantic.  “This ensures appropriate third-party review and is the standard regulated and curated process for sharing scientific research.  Myriad has published more than 8,000 variants in peer reviewed publications over just in the past two years alone, which is substantially in excess of any other laboratory.”

Initiatives like the BRCA Challenge are trying to break Myriad’s stranglehold by linking data on BRCA mutations in various public databases. “I see Starita’s research is another innovative way to tackle the problem,” adds Downing.

* This article originally stated that Myriad was currently embroiled in a legal battle over its attempt to patent the BRCA1 and BRCA2 genes. We regret the error.