Skip to main content
Premium Trial:

Request an Annual Quote

Polygenic Modifiers of Pathogenic Genes May Refine Disease Risk Prediction


NEW YORK – Researchers at Massachusetts General Hospital, Color, and elsewhere have determined that polygenic factors can significantly modify the risk conferred by monogenic risk variants for breast cancer, colon cancer, or coronary artery disease, and are exploring the best ways to implement a combined monogenic/polygenic risk score into patient care.

In a study posted late last week on the health sciences preprint server MedRxiv, first author and Mass General postdoc Akl Fadel and his colleagues sought to clarify how monogenic and polygenic risk interact by performing genetic analyses for familial hypercholesterolemia, hereditary breast and ovarian cancer, and Lynch syndrome predisposing to colorectal cancer.

For coronary artery disease, the researchers looked at a cohort of 6,449 cases and 6,430 controls derived from the UK Biobank, a prospective cohort study of middle-aged adult participants. They followed up these results in an independent cohort of 49,738 participants from the UK Biobank who previously underwent exome sequencing. For breast cancer, they studied an all-female cohort of 1,920 cases and 17,344 controls from Color, and then followed up in 27,144 female UK Biobank participants. For colorectal cancer, they studied 49,738 UK Biobank participants.

In all three conditions, the researchers found that the presence of polygenic risk variants strongly modified the risk conferred by the presence of the monogenic risk variant. For example, the presence of a monogenic familial hypercholesterolemia variant conferred a 3.17-fold increased risk of coronary artery disease in a given person. But depending on whether the individual also had a low, intermediate, or high polygenic risk score, the risk among mutation carriers for coronary artery disease ranged from 1.30-fold higher to 12.59-fold higher.

"Modeling the probability of disease by age 75 years indicated striking gradients in risk, ranging from 4.8 percent for individuals who were both noncarriers [of the monogenic variant] and in the lowest percentile of the polygenic score to 77.7 percent for individuals with a monogenic risk variant who were also in the highest polygenic score percentile," the researchers wrote.

The same held true for the breast and colon cancer cohorts. For breast cancer, the risk among monogenic variant carriers ranged from 12.8 percent to 75.5 percent, depending on whether they also had a low, intermediate, or high polygenic risk score, and risk among noncarriers ranged from 3.4 percent to 29.7 percent. Absolute risk of colorectal cancer by age 75 ranged from 11.6 percent to 79.5 percent for carriers of monogenic variants and 0.7 percent to 8.7 percent for noncarriers, depending on their polygenic risk scores.

The general idea that polygenic factors influence monogenic risk of disease isn't entirely surprising, noted Baylor College of Medicine Professor and medical geneticist Sharon Plon. Several studies have shown that common SNPs modify BRCA1 or BRCA2 risk for breast and ovarian cancer. And at the end of October, Geisinger Health System researcher David Ledbetter and his colleagues published a study in Nature Communications showing that common, polygenic factors of relevant complex traits frequently contribute to variable expressivity of rare genetic disorders, and that polygenic scores could be useful in determining the severity of those disorders.

Rather, what was unique about this new study is that even with the lowest polygenetic risk score, there was still a very clear separation between people who carried background mutations in addition to the pathogenic variants, and those who didn't. The polygenic risk was an "additive effect," she said. "Their use of the term 'modifier' is a good one. It's not eliminating the risk — it does modify the absolute risk. Your absolute risk of cancer is lower if you're in the lowest group and it's definitely higher if you're in the highest group."

Indeed, said senior author and Mass General researcher Amit Khera, the study's findings present a new way of discussing disease risk. "Right now, the traditional model is to say that if you have a BRCA1 variant, you're at high risk. And if you don't have it, you're at normal risk," he said. "What we're proposing is a new model saying if you have a BRCA1 variant, but you actually have a really good genetic background, your risk is actually pretty close to normal. It's not really significantly elevated. If you have that same mutation and there's an unfavorable genetic background, then it's a double whammy and your risk is extremely high."

Importantly, Khera and his colleagues are looking to translate the findings in this paper into the clinic as soon as possible. There is work that needs to be done to fine-tune the polygenic scores and, in some cases, understand the biological underpinnings of why the polygenic variants modify risk conferred by the monogenic variants. But while that research is being conducted, Khera is confident enough in the underlying findings to begin looking at the next steps in clinical implementation.

Now that the fundamental concept of integrating polygenic background and monogenic risk into a single model has been proven, the next step for using that model in the clinic is finding the best ways to educate doctors about how it works and show patients what it says.

Khera led the development of the recently opened Preventive Genomics Clinic at Mass General, the primary purpose of which is to implement genomic discoveries into the clinic, rather than asking primary care doctors to figure it out. The team, including Chief Genomics Office Heidi Rehm, consists of genetic counselors, laboratory geneticists, research coordinators, and physicians who are working to determine how to use genetic risk scores to improve patient care.

"Maybe [patients] have a [genetic] report from somewhere and they want to know how to put it into context. We'll help them do that," Khera said. "Other people may say, 'I'm concerned about breast cancer,' or 'What are the pros and cons of actually getting one of these tests?' We want to [help them with] that as well."

In addition, the team at the clinic is now looking at what it will take to translate the findings of the paper by Fadel et al. into clinical practice.

"How do we actually disclose it on a piece of paper or with some data visualization so that patients and doctors will understand? Because doctors haven't really heard of this to a large extent," Khera said. "So, that's one active area of research. We're developing different types of disclosure forms and conducting user testing starting this month to get feedback on how best to disclose it."

After that, he and his team plan to begin a series of pilot studies which would involve getting doctors and patients together with this new report and then using that experience to again refine how that combined monogenic/polygenic risk is being disclosed and discussed.

"The models will get better over time. The science is there," Khera said. "What I have now turned a lot of my attention to is thinking about how we might really start to move the needle in clinical care. And I think it will require some new disclosure tools, some new clinical infrastructure to really start to move forward."

In fact, Plon said, it's surprising that polygenic risk scores haven't already been added to monogenic variant risk information for certain diseases.

"Much of the early work on particular polymorphisms modifying cancer risk has been available for a long time, so I don't actually understand why it was never introduced into clinical care," she said. "There were genetic testing laboratories that could have easily added those SNPs. They never did. I think they felt that there perhaps was not enough epidemiologic data or prospective data to do that."

Though she isn't certain if this one study contains enough information to warrant an immediate change in clinical practice, she does believe that as more data becomes available about the factors that modify the risk of disease, it should be used to provide more precise information to patients.

Specifically, Plon noted, this kind of combined scoring may be most useful for patients who end up in what Khera referred to as the "double-whammy" risk group — those with a monogenic variant and a high polygenic risk score. For a woman with a BRCA1 or BRCA2 pathogenic variant, having the additional factor of the high polygenic risk score may help her decide whether to have a prophylactic mastectomy, which some women with pathogenic variants are now opting for. In the case of colon cancer, it might help a patient understand why more frequent colonoscopies are necessary, and may provide an incentive to stick to a regular surveillance schedule, Plon added.

"An important point is that the current [breast cancer] guidelines do not recommend mastectomy. The guidelines suggest it as an option for a woman who carries a BRCA1 or BRCA2 mutation to consider. And so, it certainly may be possible that, by giving someone a more appropriate breast cancer risk based not just on the mutation but on a polygenic risk score, that might cause them to make that decision differently," she said.

Khera went a step further. He believes there may also be an opportunity to change cancer surveillance guidelines in the long term.

"Whether or not a 50-year-old woman should get a mammogram screening is actually incredibly controversial, because you actually do reduce risk of breast cancer, but you induce a ton of anxiety," he said. "Many people get unnecessary biopsies and so forth. So, at present it basically comes down to a shared decision [by] the doctor and the woman. I think in the future, we'll have a more refined risk tool where you could say, 'Well, your genetics really put you at double or triple the risk of everyone else based on a polygenic score.' And that could motivate people to say, 'OK, well, in that case, I probably should get it because my risk for breast cancer at 50 is equivalent of an average person with average risk when they're 70.'"

At this point, there are still some open questions. For one thing, although the monogenic variants act to increase the risk of disease through a specific pathway, the polygenic variants appeared to affect a diverse set of physiological processes. As of yet, the biological underpinnings of these risk relationships are unexplained. But understanding how and why seemingly unrelated variants affect a patient's risk for breast cancer or coronary artery disease presents an important opportunity for research and for the development of new drugs, Khera said.

"If you look at the genetic drivers of heart attack, we find that about 20 percent of them are related to the cholesterol pathway, and we have a bunch of drugs for that. But when we look at the remaining 80 percent, 30 percent [of them are in] pathways like blood pressure. But the other 50 percent of the genetic drivers for a heart attack, we actually don't know what they do or why they do it," he added. "We're definitely looking to clarify those biological mechanisms, and ultimately the goal of that is to find new pathways that could be altered for drug therapies."

The authors also acknowledged that information concerning monogenic risk variants in general, and the development of the polygenic scores, has been based primarily on patients of European ancestry, which affects the utility of these scores for patients of other ancestries. Plon noted that she doesn't believe the data is ready to be used in patients who aren't of European ancestry, and that the medical genetics community should wait for the results of ongoing efforts to develop ancestry-specific polygenic risk scores before applying those scores in patients who aren't Caucasian.

Khera is a bit more optimistic. Although he noted that his group is continuing to conduct research on diverse cohorts of patients, he also said the basic concept of combining monogenic and polygenic risk scores does work to predict risk across ethnicities, though it currently performs best in individuals of European ancestry.

The issue isn't that the underlying biology is necessarily different, he noted. But because there are more people of European descent than people of other ethnicities in large genomic databases, it can be harder to gather the same numbers of sample sizes.

"We're working actively to collect much more diverse participants," he said. "It already predicts risk in every ethnicity. But I think it's going to level out that playing field in the next couple of years. I don't think the biology is different — I think this concept is very common, and it holds true whether you're Chinese, or black, or Indian, or white. But just how good these polygenic scores are does vary a little bit currently by ancestry. So, that's something we're working actively on."