By Aaron Wolf, Ph.D., Science Communications at Ancestry
Your AncestryDNA Traits results are based on your DNA. By looking at your DNA, we can make predictions about your appearance, sensitivity to certain tastes, whether you'll remember your dreams, and more.
We currently have two ways of developing our traits predictions. One of these ways is to use polygenic risk scores. Polygenic risk score (PRS) traits look at hundreds or thousands of DNA markers and how they relate to other members’ responses to our survey questions. Our scientists then build a mathematical model to predict a person’s trait based on their DNA at these markers.
AncestryDNA Surveys help us research new PRS traits. So when you answer survey questions, you help us build more PRS traits for you and other AncestryDNA users. In the end, though, your trait results are based on your DNA—not your answers to survey questions.
How we build PRS traits
PRS traits use responses to survey questions (answered by people who have consented to participate in research) to connect thousands of DNA markers to a trait.
Consenting AncestryDNA users answer survey questions and contribute their DNA data to help build the PRS trait prediction models.
Survey responses and genetic information are combined to identify a set of DNA markers that are statistically associated with the trait. This type of analysis is called a genome-wide association study. For a single trait—like your ability to remember dreams—these analyses may identify hundreds to several thousands of DNA markers associated with that trait. (Our scientists discovered over 7,200 genetic markers associated with remembering dreams.)
A Manhattan plot shows the several thousands of DNA markers (represented as dots) that show a statistical association with the trait Remembering Dreams. They were used to build the trait prediction model. The higher a dot is in the plot, the stronger the DNA marker’s association with the trait.
Genome-wide association studies can also indicate the effect size, or importance, of each DNA marker in predicting that trait. While there may be thousands of DNA markers associated with a trait, many of them have a minimal effect. In this way, each marker provides a tiny “nudge” in the direction of the trait.
It’s also true that not all relevant DNA markers may point in the same direction for an individual. For example, someone may have several hundred markers that make them less likely to remember their dreams and another several hundred that make them more likely to remember their dreams.
By adding together all the genetic markers associated with a trait and their effect sizes, our scientists can make a prediction for how likely you are to have that trait.
PRS traits compare you to others
You may have noticed in your AncestryDNA Traits report that the results for PRS traits, like Remembering Dreams or Risk Taking, are presented relative to other AncestryDNA users who answered the survey question.
Example report for Remembering Dreams trait
This is because a PRS can only provide a relative risk score—how a person’s likelihood for a trait compares to others with different genetic makeup. Each polygenic score can be put on a bell curve distribution. Most people will find their scores in the middle, indicating an average likelihood compared to the rest of the population.
The relative risk scores for a PRS trait follow a bell curve distribution. Everyone’s score is somewhere on the distribution.
Importantly, the PRS doesn’t provide a baseline probability for the trait or the absolute risk. Often, large differences in relative risk do not correspond to a large difference in absolute risk. This is especially true when the absolute risk for a trait is very low. For example, say the absolute probability of being born on Mars is 1 in 1,000. Even if the relative probability of being born on Mars is twice that for children of astronauts compared to children of non-astronauts, the absolute probability of being born on Mars for children of astronauts is only 2 in 1,000. That’s still a very low absolute probability.
Genetics is not destiny
While a PRS looks at a lot of genetic information to make a prediction about a trait, there’s a lot of non-genetic information it doesn’t consider. Environmental factors, life choices, and random chance can all impact whether you do or don’t have a trait, regardless of the likelihood predicted from your genetics. Even traits with a very strong genetic component, like height, are strongly influenced by non-genetic factors like early childhood nutrition.
Building PRS traits for the future
Ancestry scientists are committed to developing new and exciting trait predictions. Because building the PRS prediction models requires combining information from user survey results and DNA results, the power of our predictions relies on the data we get from users.
You can help us build new PRS traits by completing our DNA Surveys and consenting to participate in our scientific research.
Still curious to understand more?
Cool--we're glad you're as interested in Traits as we are. Check out our white paper on Traits.