Percentiles, Populations, and Distributions. Oh My! Genetic Scoring for Athletic Ability at Athletigen

Because this is an area where interpretation of the results can be a little more nuanced, I wanted to talk a little bit about percentiles as they apply to our athletic trait scoring, reasons why they may be low (but not mean you aren’t a great athlete), and some general take always

1-cJeeWvhEJRZX178H-FlRbw (1) (1)

My Motivation Score Expressed as a Percentile Placed on a Distribution

For motivation, I’m in the lower 37th percentile, which means that most people in our database have a higher motivation score than me. But the number alone doesn’t tell the whole story. To me, this percentile makes sense, I do struggle with the motivation to get off my butt and go work out. It isn’t that I don’t enjoy being physically active, I do, and as a kid and teenager I was quite active playing Hockey (like most Canadian boys) and generally running around outside. But it is easy for other things to take priority for me, and as an adult I spend far less time being active than I should. And that is what the markers we have in our database related to Motivation measure. In most cases they have been directly correlated with how frequently people are active, and how high that activity level is (low impact, moderate, high impact).

We can also see that where people fall on the scale of motivation scores isn’t uniform. The distribution of scores (where the y-axis here can be thought of as the number of people and the x-axis is the scores from 0 to 1o0) is lumpy. I’m just on the lower side of a big peak of individuals with scores fairly similar to mine. One or two point scores in either direction would see a lot of movement in my percentile score. In this distribution there is another nearby peak of people who are more physically active than most of the population, and a smaller peak at the upper end of things who are very active.

The other issue to keep in mind is that what we are looking at (or calculating) are essentially genetic scores for favourability. While in some cases the opposite of the favourable allele may be detrimental, in many cases it is phenotypically ‘normal’ and not an impaired state. Motivation may not be the best example of this though, since the “normal” phenotype may be correlated with low (bad) activity levels. But something like endurance probably is:


1-lTbWuDk3KuuMOeXydKeoNA (1)

In endurance we have a somewhat normally distributed set of scores but there is a slight (but distinct) skew to the lower range. Most people people in general don’t have a lot of markers associated with enhanced endurance performance. But scoring on the lower end of the distribution just means you are relatively ‘normal’ for endurance capacity in terms of your total genetic contribution.

The other issue here, is that many of the genetics that have been examined regarding athletic potential have been looked at statistically. Do you observe certain alleles at greater frequencies among elite athletes than you do in the general population? This means that for most markers there are lots of people who are non-athletes who have great genetics. There are other reasons why they didn’t go that route. There are also plenty of currently unknown genetic markers affecting athletic ability that we will discover as we move forward with Athletigen.

The main take aways regarding these scores is that they measure a simple additive or aggregate score of what your genes contribute to performance based on some of the better researched markers associated with enhanced athletic performance. Your percentile is a reflection of where you fit among the general population in terms of this enhanced performance capability, and because of the nature of the scores small changes in scores can result in larger or smaller percentile gains depending on the specific distribution to a category.

We have also learned a few things ourselves based on the feedback we have been getting. One is that the percentiles alone can be difficult to interpret and we didn’t provide enough context for allow users to do that very well. At least giving you your raw genetic score out of some uniform scale (1–100) would have helped quite a bit. And really the most important thing people are interested in are specific genetic markers and what they mean.

And so we are moving towards that in an upcoming release. Focusing more on just showing you individual markers that have a high confidence associated with them, how they effect your athletic potential, and what actionables may exist in order to help you improve your training. We will still keep the scoring but we will be adding additional context to the scores and moving towards giving you multiple ways of viewing that data.

%d bloggers like this: