Penn scientists have found a way to separate the influence of genetic and environmental factors on disease risk.

Every disease is shaped by a genetic component, as well as environmental factors such as air pollution, climate, and socioeconomic status. However, the extent to which genetics or environment play a role in disease risk -; and how much of each can be attributed -; is not well understood. As a result, the actions individuals can take to reduce their risk of disease are often not clear.

A team led by researchers at Penn State College of Medicine has found a way to distinguish between genetic and environmental effects of disease risk using a enormous, nationally representative sample. They found that in some cases, previous estimates overestimated the contribution of genes to disease risk, and that lifestyle and environmental factors play a larger role than previously thought. Unlike genetics, environmental factors such as exposure to air pollution can be more easily modified. That means there are potentially more ways to mitigate disease risk. The researchers published their work in Nature communication.

We are trying to disentangle the extent to which genetics and environment influence disease development. If we better understand how each contributes, we can better predict disease risk and design more effective interventions, especially in the era of precision medicine.”

Bibo Jiang, assistant professor of public health sciences at Penn State College of Medicine and senior author of the study

The researchers said environmental risk factors have historically been tough to quantify and measure because they can include everything from diet and exercise to climate. However, if environmental factors are not considered in disease risk models, the analyses can wrongly attribute shared disease risk among family members to genetics.

“People living in the same neighborhood share the same levels of air pollution, socioeconomic status, access to health care providers, and food environments,” said Dajiang Liu, distinguished professor, vice chair for research, director of artificial intelligence and biomedical informatics at Penn State College of Medicine and co-senior author of the study. “If we can separate out these shared environments, what remains may more accurately reflect the genetic heritability of disease.”

In this study, the team developed a spatial mixed linear model (SMILE) that includes both genetic and geolocation data. Geolocation—an individual’s approximate geographic location—served as a proxy for community-level environmental risk factors.

Using data from IBM MarketScan, a health insurance claims database with electronic health records of more than 50 million people with employment-based insurance policies in the United States, the research team filtered information on more than 257,000 nuclear families and compiled treatment outcomes for 1,083 diseases. They then supplemented the data with publicly available environmental data, including climate and sociodemographic data, as well as levels of particulate matter 2.5 (PM2.5) and nitrogen dioxide (NO2).

The team’s analysis led to more refined estimates of the factors that contribute to the risk of the disease. For example, previous studies had shown that genetics contributed 37.7% to the risk of developing type 2 diabetes. When the research team re-evaluated the data, their model, which took into account environmental influences, showed that the estimated genetic contribution to the risk of type 2 diabetes decreased to 28.4%; most of the risk of the disease could be attributed to environmental factors. Similarly, the estimated contribution to the risk of obesity attributable to genetics decreased from 53.1% to 46.3% after taking into account environmental factors.

“Previous studies have shown that genetics played a much larger role in predicting disease risk, and our study recalibrates those numbers,” Liu said. “That means people can remain hopeful even if they have relatives with type 2 diabetes, for example, because there is a lot they can do to reduce their own risk.”

The research team also used the data to quantify whether two specific air pollutants — PM2.5 and NO2 — had a causal effect on disease risk. Previous studies, the researchers say, grouped PM2.5 and NO2 together as a single aggregate measure of air pollution. But in this study, they found that the two pollutants had distinct and separate causal relationships with health conditions. For example, NO2 has been shown to directly cause conditions such as high cholesterol, irritable bowel syndrome, and type 1 and type 2 diabetes, but not PM2.5. On the other hand, PM2.5 may have a more direct causal effect on lung function and sleep disorders.

Ultimately, the researchers say, the model will allow for deeper analysis of questions about why certain diseases are more common in certain geographic locations.

Other Penn State authors on the work include Havell Markus and Austin Montgomery, dual medical and doctoral students at Penn State College of Medicine; Laura Carrel, professor of biochemistry and molecular biology; Arthur Berg, professor of public health sciences; and Qunhua Li, professor of statistics. Daniel McGuire, a doctoral student in biostatistics at the time of the study, co-led the study. Co-author Lina Yang and Jingyu Xu, who were doctoral students in biostatistics at the time of the study, also contributed to the paper.

This work was supported in part by the National Institutes of Health and the Penn State College of Medicine Artificial Intelligence and Biomedical Informatics Pilot Funding Program. Some materials used in this work were provided by the Center for Applied Studies in Health Economics at Penn State College of Medicine.