For more than a decade, scientists with the PAGE consortium (Population Architecture using Genomics and Epidemiology) have been focused on developing polygenic risk scores — those statistical estimates of a person’s risk for disease based on their genes — that will accurately predict risk in all major racial and ethnic groups in the U.S.
This is crucial since the Human Genome Project — and a lot of the of genetic research done since — has focused almost exclusively on people of European ancestry. As a result, polygenic risk scores designed to help clinicians and the general population gauge their susceptibility to various diseases aren’t as accurate as they could be, particularly when it comes to groups with non-European ancestry, such as African Americans, Latinos and Indigenous populations.
PAGE researchers at Fred Hutchinson Cancer Research Center, including biostatistician Dr. Charles Kooperberg, molecular geneticist Dr. Ulrike “Riki” Peters and genetic epidemiologist Dr. Chris Carlson, have already started using the strength of large numbers — population science’s secret weapon — to analyze the genetic variants of different ethnic groups in order to create accurate, unbiased polygenic risk scores for colorectal cancer.
Now, they’re doing the same thing in cardiovascular disease, thanks to a $9.8 million grant from the National Heart, Lung, and Blood Institute, one of the 27 institutes and centers that make up the National Institutes of Health.
“The goal is to create risk scores for cardiovascular and related diseases that work equally well in people of color as in European Americans,” said Kooperberg, head of the Hutch’s Biostatistics Program and one of the principal investigators of the study. “Our ultimate goal is to reduce and prevent the burden of cardiometabolic diseases in all populations.”
Polygenic risk scores are already widely used in commercialized DNA tests produced and marketed by Ancestry, 23andMe, Myriad and many other outfits, with both good and bad outcomes.
But they’re just now starting to be used by clinicians to help people understand how family history, behavior and environmental risks can come together to inform their susceptibility for certain diseases like cancer or cardiovascular disease.
“Risk scores will be of good use in personalized predictive medicine,” Kooperberg said.
But only if they’re accurate. Right now, the strong Eurocentric bias means they’re underperforming for everybody, especially when it comes to those with mixed ancestral backgrounds.
Peters said PAGE was the first group to demonstrate with empirical data that genetic risk prediction underperformed in people of color. A flagship paper, published by the group in the journal Nature last year, further strengthened the argument that findings from one population cannot always be generalized to others.
“Polygenic risk prediction is moving more and more into clinical use, but we are not as good at predicting people’s risk if they’re non-Europeans,” Peters said. “There are people out there who will be able to benefit from preventive measures if they’re better identified as high risk. That’s why there’s an urgency to fix this.”
Another compelling motivation: Current scores used to stratify and treat patients could potentially be doing harm.
“You might overtreat people who don’t need treatment or undertreat people at the highest risk,” Kooperberg said.
Creating unbiased polygenic risk scores should fix the issue and ensure that people of color “are not the last to benefit in the new era of genomic medicine,” the researchers said.
The diseases that plague humans are just as diverse as we are. Some, like cystic fibrosis, are caused by a single gene variant, or mutation. Others, like cardiovascular diseases, can be caused by multiple variants.
Most coronary artery disease, a type of cardiovascular disease, is triggered by many different genetic variants scattered across the genome. These variants often work hand in glove with other genetic changes driven by environmental influences, both positive and negative.
Some of these factors — whether you can access clean air, water and fresh food; whether you smoke, drink, exercise or eat a healthy diet — are malleable. People may be able to limit exposure to some of these risk factors in an effort to prevent disease. Others are fixed, but can be improved by medications, like statins to lower high cholesterol.
By analyzing the genetic data of large subsets of patients and pairing it with their outcomes data, PAGE researchers will be able to aggregate all manner of genetic risk variants — inherited and environmental — into a single risk score, then stratify people into different risk categories on a scale.
The more data they have, the bigger the scale and the clearer the score.
“You can get a score, a continuous number,” Kooperberg said. “Then you can look at the lowest 10%, the next 10% and so on and so on. You can see what the risk is from the top decile to the bottom decile. In an ideal world — we’re not there yet — you can identify the people who are more or less likely to get heart disease. The top group, with the highest risk, would be the ones you’ll monitor and put on statins or other preventive measures, such as physical activity interventions.”
This last phase of the ongoing PAGE project will involve the genetic data of 1.5 million participants with non-European ancestry and a dozen collaborating institutions, with the Hutch serving as the data coordinating center. Principal investigators include Kooperberg, Dr. Christopher Gignoux of the University of Colorado Boulder and Dr. Kari North of the University of North Carolina at Chapel Hill. Peters is also part of the leadership team.
Collaborating centers will provide patient data gleaned from two dozen diverse cohorts and biobanks, together with linked electronic health records that capture a diagnosis of cardiovascular disease, along with its risk factors and outcomes.
“A lot of these large studies have already done genotypes and sequencing,” Kooperberg said. “It’s merging the data from various cohorts.”
Hutch researchers will then use a machine learning approach to analyze the data, then develop polygenic risk scores for cardiovascular disease-associated traits in racially/ethnically diverse populations, validating those scores against real patient data and outcomes.
“There are millions and millions of genetic variations in the genome and many of them may slightly increase risk for cardiovascular disease,” Peters said. “A few carry larger risk, many more carry small risk. When you bring them all together, you can start building risk scores that have sufficient predictive power.”
Peters said many of the variants may still be unidentified, but with machine learning approaches “even if you don’t have them identified, we can use all the genetic information and give them some predictive weight.”
With this approach, she said, “we can improve the prediction and develop the best-performing polygenic risk score.”
Once the risk scores are validated, the researchers will begin the process of creating infrastructure and best practices to disseminate this information out to clinics.
The PAGE group, which Kooperberg termed “a close, collaborative family,” has been working on genetic research for populations of diverse ancestry since its formation in 2008.
“We recognized early on that a lot of the emphasis in genome-wide association studies was on people of European ancestry, and that didn’t seem right,” he said. “When you’re testing 100 samples to see if a technology works, you use the samples you have access to, but we are well past that point now.”
Kooperberg said he hopes their work will help reduce the health disparities that exist in diverse populations. He also said he’s thrilled to see the PAGE consortium’s vision finally come to fruition.
“We are a little bit ahead of the curve, but not far ahead,” he said. “But this is exciting. This is good. We’re making progress and I’m confident we’ll continue to make progress.”
Diane Mapes is a staff writer at Fred Hutchinson Cancer Research Center. She has written extensively about health issues for NBC News, TODAY, CNN, MSN, Seattle Magazine and other publications. A breast cancer survivor, she blogs at doublewhammied.com and tweets @double_whammied. Email her at firstname.lastname@example.org. Just diagnosed and need information and resources? Check out our patient treatment and support page.