Quantcast
Channel: io9
Viewing all articles
Browse latest Browse all 36042

What Heritability Looks Like

$
0
0

I am talking to people looking for actual geneticist to do actual genetics work. This is very strange. Genetics, the study of inheritance, is rarely done anymore. This brings me to a problem of how I see heritability of traits portrayed in both popular media and biologists and clinicians who are not geneticists.

What is heritability? It is a test statistic like a t-value in a t-test or an F-value in ANOVA. It represents nothing real or tangible except in relation to other heritability measures. As heritability increases, the degree of influence of inheritance is presumed to increase.

What Heritability Looks Like

Heritability is often misused. If a study says “height is 80% heritable,” some people will claim that your “genes determine 80% of your height.” This is not true. What it actually means is that genetic factors are estimated to determine 80% of the value of the heritability test statistic. This means any heritabilities measured at under 80% are probably less influenced by genetics than height is and any that are greater than 80%... well, we will get to that shortly.

Can I Compare Heritabilities I Read About in Genetics for Non-Majors to the Latest Study I Read About on BBC News?

No, you cannot. It is important to understand that heritability is only meaningful when compared to other measures of heritability that are calculated the same way and, preferably, in the same population or even the same study.

There are different ways to measure heritability. Lately the favored method is ANOVA, but previously studies used a simpler mathematical model that required special conditions to work in human studies. They give you different values with the same data and only provide meaningful comparisons if heritability values are grossly different (e.g., 30% v. 70%, but not 40% v. 50%).

Wait, Why Can’t I Compare Heritabilities of the Same Trait Measured in Burma and Sweden?

Heritability, as a test statistic, is influenced by environment. The actual contribution of inheritance is less influenced by environment. The sequence of the genes themselves is not influenced by environment (with some caveats to that). Let’s take specific example: some people claim homosexuality is influenced by genetic factors.

Let’s imagine a society where gay men are completely accepted but lesbians are considered utter abominations to be cast out, persecuted and even murdered. In that case, you expect some men who have a weak preference for other men to be gay, but women will only be lesbian if they absolutely cannot suppress their desire for other women. In this case, if there is any genetic influence it will be hard to detect above background variance in gay men. Lesbianism will appear much more highly heritable because the only lesbians will be those where all or nearly all the factors influencing sexual orientation (whatever they are) are pushing them towards preferring women.

There are actual formulas to estimate the influence of specific alleles of specific genes on a trait. They can get pretty complicated. Read up on the writings of geneticists like Lande if you are interested in that.

You Said Something About Heritabilities Over 80%?

Heritability is a test statistic. It is a measurement and, like all measurements, has limitations. If you have a scale designed to measure weights of small children that only goes up to 60 pounds, an adult man and a gorilla will give the same measurement: 60 pounds. Heritability has a more subtle limitation.

Heritability is most often measured by ANOVA these days and ANOVA works by a mathematical technique called partitioning sum of squares. I will not get into the mathematics of it here, but suffices to say if there is any mean difference between two groups, even if the difference is by chance alone, some of the sum of squares will be “explained” by the statistical model. This means heritability values of 0 or 100% will never happen, even if a trait is not influenced by inheritance at all by or is completely governed by inheritance. Let me show you:

What Heritability Looks Like

Suppose the actual heritability of a trait is 40%. Different studies might get different measures such as 42% of 31%. If you perform an infinite number of studies, the heritaiblities will average 40%. If the trait shows little variance (like the red curve), the studies will all give heritabilities close to 40%. With higher variance traits, two studies might give widely varying heritabilities such as 26% and 50%.

However, if heritability is very high or very low, strange things occur. Heritability measures cannot go above 100% or below 0%. To use the homosexuality example, “negative homosexuals” and “negative heterosexuals” do not exist. Because any 2 groups in the sample will have different proportions of heterosexuals and homosexuals, a heritability will be measured even if inheritance has no effect at all.

What do those data look like?

What Heritability Looks Like

You will not get a nice Gaussian curve around the actual value (in this case, 0). You will always or nearly always measure something above zero. For a high variance trait, this may be far above zero (like 15%), while a trait with low variance will be closer to zero (like 9%). As a general rule (with some exceptions), I do not believe any heritability below about 20% has any influence of inheritance and any heritability above 80% is significantly influenced by other factors. Granted, I have almost never seen a published heritability over 80%, but I have seen many below 20%.

Why Do You Keep Saying Inheritance Instead of of Genes?

Great question! That is a topic for another day, my friend.


Viewing all articles
Browse latest Browse all 36042

Trending Articles