NEW YORK —Researchers from population genomics company Helix have used exome sequencing and medical record data to uncover several conditions that could be included in population genetic screening programs to uncover additional individuals at risk of disease.
Conditions used in population genetic screening programs are generally common, have a genetic component with high penetrance, and are actionable. Currently, there are three genetic conditions designated "tier one" by the US Centers for Disease Control and Prevention: BRCA-related hereditary breast and ovarian cancer, Lynch syndrome, and familial hypercholesterolemia. Screening an unselected population for these conditions identifies a pathogenic or likely pathogenic variant in about 1 percent of individuals, most of whom — about 80 percent — are unaware of their increased disease risk.
Using exome sequencing and medical record data from more than 220,000 participants, researchers from Helix sought to identify additional conditions that might meet these criteria. As they reported in Genetics in Medicine last week, they uncovered seven genetic conditions, four of which are not currently used in population genetic screening, that could identify additional individuals with actionable genetic risk.
The goal of the project was to find new conditions, beyond the three that the CDC has identified, that could have a role in broader population genetic screening, said study co-author William Lee, VP of science at Helix.
The approach taken by the Helix team addresses an outstanding problem in how genes are selected for population screening, said Adam Buchanan, director of genomic medicine at Geisinger and co-director of its MyCode Genomic Screening and Counseling program, who was not involved in the study. He noted that genetic variants in lists of secondary findings like the one from the American College of Medical Genetics and Genomics can have very low penetrance in an unselected population and relying on them could subject individuals to unnecessary procedures.
"I think starting with a larger dataset — electronic health record or patient report data and the genomic data as well — is a great starting point," he said.
In their analysis, Lee and his colleagues conducted gene-disease association analyses using two cohorts — 189,495 individuals from the UK Biobank and 28,423 individuals from the Healthy Nevada Project — that underwent exome sequencing and phenotyping. Through their gene-based collapsing analysis, they uncovered 74 statistically significant associations between 27 genes and 49 phecodes, or groups of related phenotypes drawn from medical record billing information.
The researchers also applied a percent predictive value cutoff of 0.3 to these associations. That is, at least 30 percent of individuals with variants in those genes would have to develop the associated condition. Current gene-disease associations included in population genetic screening have PPVs of about 0.3, the researchers noted, and by applying that cutoff value to their new set of genes, they homed in on seven genes they suspected would have utility.
This approach, Buchanan added, increases the likelihood that someone with the variant might develop the disease and addresses the concern that people could be subjected to unnecessary or risky procedures. He added that other screening approaches, like mammograms and colonoscopies, have PPVs lower than the cutoff used in this analysis, though there is not one standard, accepted cutoff value.
Three of the associations the Helix team identified — linking variants in BRCA1 and BRCA2 with breast cancer and variants in LDLR with coronary atherosclerosis — are already used in population genetic screening efforts and functioned as a sort of positive control for the analysis.
But four other associations — between loss-of-function variants in HBB and hemoglobinopathies, loss-of-function variants in PKD1 and cystic kidney disease, coding variants in GCK and diabetes mellitus, and coding variants in MIP and cataracts — are not currently in use in screening.
Buchanan noted that how the Helix team identified these gene-disease links differs from how Geisinger's MyCode effort generally determines which genes to include. "It's really intriguing," he said, noting that Geisinger has access to UK Biobank data and its own large cohort of individuals with which they could do validation studies of the Helix results.
Adding variants in these genes to genetic population screening efforts could identify additional individuals with actionable disease risk, the researchers wrote. For instance, variants in PKD1 and PKD2 cause autosomal dominant polycystic kidney disease. While there is no cure for this condition, detecting it early could prioritize patients for surveillance of total kidney volume and estimated glomerular filtration rate. Further, early detection could help treat co-morbidities like early-onset hypertension, cardiovascular complications, and cyst infections, as well as potentially slow disease progression through pharmaceutical interventions.
"Often kidney disease isn't caught until the much later stages when it can often be very difficult to treat," said co-author Nicole Washington, director of research at Helix. "There's a lot of other symptoms that the patients experience, so identifying it early would behoove us to improve their outcomes."
Folding these four genes into current population genetic screening programs would uncover an additional 0.21 percent of participants with actionable disease risk, according to the researchers' analysis, for a total 1.21 percent of an unselected, screened population. "[This] doesn't sound like a lot," Lee noted. "But if you're screening hundreds of thousands or millions of people, it does turn out to be significant."
In addition to further validation, Lee noted that new infrastructure might be needed to adopt these new genes clinically. Some of the diseases fall under medical specialties where genetic screening is not yet used and that might not have the foundations in place to support it. "One of the goals of doing the study to begin with was to, at least, get the ball rolling so that we could roll these out in a responsible way," he said.
Lee added that he and his colleagues plan to keep analyzing new data as it is added to those datasets as well as analyze cohorts with more diverse ancestry — both the UKB and HNP cohorts used in the Helix analysis predominantly included individuals of European ancestry. He noted, though, that because of the stringent cutoffs used, the genes they linked to conditions are not likely to be specific to a particular human subpopulation.
"I think over time, you can imagine being able to identify additional such conditions, maybe some of which would be specific to some subpopulations as you get more data," he added.