Skip to main content
Premium Trial:

Request an Annual Quote

ARUP Takes on BRCA Variant Classification by Launching Database, In Silico VUS Resource


NEW YORK (GenomeWeb) – Labs conducting BRCA testing have yet another resource when they encounter a variant with unclear links to cancer, this time housed at ARUP Laboratories.

The University of Utah, along with its Huntsman Cancer Institute and its non-profit lab ARUP, announced last week the creation of their own open-source BRCA1 and BRCA2 gene mutation database. ARUP's database is one of many options that researchers, genetic counselors, and oncologists can use to learn about whether alterations in these two highly variable BRCA genes are associated with hereditary breast and ovarian cancer.

Quest recently announced it would fund curation and functional studies to enhance BRCA1/2 variant classifications within the Universal Mutation Database (UMD), which is housed at Inserm, France's national institute of health and medical research. Laboratory Corporation of America has said it will join Quest's BRCA Share effort, which will be free for patients and researchers, but commercial labs will have to pay a fee on a sliding scale.

Other publicly accessible databases for BRCA variant classification include the Breast Cancer Information Core; the Leiden Open Variation Database 2.0 (LOVD); and the Human Gene Mutation Database. But these resources have received much criticism for being poorly maintained and for containing errors.

Meanwhile, the NIH has launched a large-scale variant classification effort called ClinGen, within which there is a dedicated group focused solely on BRCA1/2, dubbed the Sharing Clinical Reports Project. But ClinGen is very much under development, and many in the life sciences field believe the database is not ready for use in the clinical setting. In this environment, Quest is hoping to fill the need for a well-curated, clinical-grade BRCA variant database by improving UMD and getting large and small labs to submit data.

In launching its own database in collaboration with the University of Utah and Huntsman Cancer Institute, ARUP is also hoping its open-source database will provide labs, researchers, and doctors key insights into BRCA variants that may not be available within other repositories. "ARUP develops and hosts databases when a locus-specific database that is regularly curated and updated isn't available," Elaine Lyon, medical director of the molecular genetics and genomics department at ARUP, told GenomeWeb. "Even though there are some BRCA1/2 databases, they may not have been curated extensively or not curated clinically."

ARUP provides two types of resources on BRCA1 and BRCA2 variants — one that classifies variants from published literature and family studies, and another that provides in silico predictions of risk to advance understanding of variants of unknown significance (VUS). The first resource contains data on around 2,500 reported BRCA1/2 variants, while the in silico database contains predictions on approximately 1,900 BRCA1 and 3,400 BRCA2 amino acids

Since the launch of BRCA testing in the mid-1990s, clinicians have observed a large number of single-nucleotide substitutions or missense mutations — as many as 10,000 by some estimates — in these two genes. However, researchers have not been able to characterize how many of these rarely seen VUS impact resulting proteins, and so, their association to cancer remains unclear. When variants have unclear or uncertain association to disease, they are called VUS.

One of the advantages to ARUP's approach, according to Lyon, is for these poorly understood VUS, one can go into the in silico database and learn about the risk probability. The prediction methods employed in that database have been largely developed by Sean Tavtigian, a professor in the oncological sciences department at the University of Utah. Lyon believes his method likely provides more accurate risk predictions than more general in silico programs.

ARUP's in silico database for BRCA1/2 provides a risk prediction for each nucleotide change. "A gene-specific in silico analysis hopefully is more accurate than general prediction programs used for any gene," Lyon said. "If you can train a prediction program for specific genes, the accuracy rate of the calls should be higher."

Tavtigian's in silico method starts with establishing a prior probability that a particular sequence variant is neutral or pathogenic in terms of risk, and uses Baysian statistics to factor in observational data and arrive at a posterior probability. In the case of BRCA1/2 VUS, the prior probability is influenced by whether the position of a variant is in the functional domain of a protein. If it is, then in silico analysis is used to weigh the impact of the missense mutation on proteins and determine if the variant will damage RNA splicing.

Observational considerations are then factored in — such as whether the variant is co-segregated, if the VUS is occurring with a clearly pathogenic variant, the tumor histopathology, as well as the family and personal medical histories of the VUS carrier — to arrive at a posterior risk probability of between 0 and 1. Experts have developed guidelines and tables that inform what labs and doctors should do based on the posterior probability estimate. For example, a high probability that the BRCA VUS is pathogenic might have medical implications for the patient, but if there is more of an intermediate probability of the variant as being pathogenic or neutral, the doctor might advise the patient to join a research study or recommend other family members be tested to gather more information on the variant.

Before any analyses are conducted on BRCA VUS, these markers average a 10 percent probability of being pathogenic, according to Tavtigian, who has described his BRCA-specific in silico method in an online talk. In comparison, VUS in mismatch repair genes associated with increased risk of colon and endometrial cancers have around a 50 percent average prior probability of pathogenicity. Given the relatively lower prior probability of pathogenicity for BRCA VUS, Tavtigian explains in this talk that while only a little bit of observational data is needed to push a VUS toward a neutral likelihood (not associated with cancer), quite a bit of data is required to move it toward having a high probability of pathogenicity.

Most of us are very willing to share what we have, because we understand that no one lab will be able to collect information at a global level.

Including its BRCA databases, ARUP now operates 10 such free, open-source variant resources for inherited diseases. ARUP formally started building its BRCA1/2 databases two years ago, after the Supreme Court invalidated a number of Myriad Genetics' patent claims underlying BRACAnalysis, Myriad's breast and ovarian cancer risk test that dominated the market for nearly two decades.

Several of the BRCA patent claims at issue in the Supreme Court case were owned by the University of Utah, which it exclusively licensed to Myriad. Meanwhile, the University of Utah also owns ARUP, and so, in launching this BRCA variant database, the parties involved had to "work within the parameters of the university's licensing agreements with Myriad," Lyon said.

With the court's decision, however, the BRCA testing market opened up and labs quickly realized that they needed to work together to address the VUS challenge. "Clinical labs collect what information they can, and most of us are very willing to share what we have, because we understand that no one lab will be able to collect information at a global level to do this," Lyon said. "We refer to this [variant classification] as pre-competitive space."

Toward this end, ARUP has submitted some data into ClinVar, the database housing variant information within NIH's data-sharing effort ClinGen. According to Lyon, before submitting to ClinVar, the data needs to be "cleaned up" and formatted. ARUP is still working on ways to make the submissions process more seamless between its own database and ClinVar.

The goal of ClinGen is to eventually develop ClinVar into a resource doctors and labs can use to inform clinical decisions. Although within ClinGen there are ways in which labs and clinicians can resolve discrepancies on variant calls, ClinVar accepts submissions from many sources, and may contain information that's outdated or not well curated, Lyon said, advising users to consider data and data sources in ClinVar carefully before using it to inform patient reports.

"I have seen submissions to ClinVar that includes classifications based on how Myriad has reported it, for example," Lyon said. "Although useful, a report from Myriad isn't the same as having published papers with evidences that we can assess ourselves."

Meanwhile, Myriad continues to operate its large variant database in a proprietary fashion and recently published a paper highlighting how infrequently publicly accessible variant repositories agree in their BRCA variant classifications. The analysis, published in the Journal of Community Genetics, retrospectively analyzed 2,017 BRCA variants, of which 116 were identified as pathogenic in at least one database. However, of these 116 variants, all the databases agreed on the classification for only four. Meanwhile, 34 percent of the mutations that Myriad identified using its own database didn't show up in any of these other repositories.

One of the positive aspects of ClinVar, in Lyon's view, is that it alerts labs that have submitted data when the classification for a variant differs between labs. "We can get on the phone and talk with the other laboratory. We may simply have different internal evidences that we are using," she explained. By sharing each lab’s information, often the labs can come to an agreement on the classification.

Lyon has found that most of the divergent calls between labs aren't glaring discrepancies, where one lab has called a marker benign and the other has called it pathogenic. Often times the discrepancies are between a likely pathogenic or pathogenic variant call. "I call that a one-degree of separation," Lyon said. "And we're starting to understand the reasons for the discrepancies. I’m very hopeful that, within the next few years, labs will be able to resolve these discrepancies."