Monday, 29 August 2016

Paper out

Our paper on genetic assortative mating in the UK is out in Intelligence . Here's the abstract:
We examined whether assortative mating for educational attainment (like marries like) can be detected in the genomes of ~1600 UK spouse pairs of European descent. Assortative mating on heritable traits like educational attainment increases the genetic variance and heritability of the trait in the population, which may increase social inequalities. We test for genetic assortative mating in the UK on educational attainment, a phenotype that is indicative of socio-economic status and has shown substantial levels of assortative mating. We use genome-wide allelic effect sizes from a large genome-wide association study on educational attainment (N ~ 300 k) to create polygenic scores that are predictive of educational attainment in our independent sample (r = 0.23, p b 2 × 1016). The polygenic scores significantly predict partners' educational outcome (r = 0.14, p = 4 × 108 and r = 0.19, p = 2 × 1014, for prediction from males to females and vice versa, respectively), and are themselves significantly correlated between spouses (r = 0.11, p = 7 × 106). Our findings provide molecular genetic evidence for genetic assortative mating on education in the UK.

So, why's this interesting? Here's a brief summary of the social science of individual behaviour for the past 20 years or so:
SOCIAL SCIENTIST: There are correlations between parents' and children's behaviour! This proves that upbringing and social structure matter.
GENETICISTS: Dude! It's probably just shared genes.
So, lots of things seem to be handed down from parents to children: wealth and income. Moral attitudes. Which political party you support. But then, twin studies would show that monozygotic twins (identical, all the same DNA) were much more similar than dizygotic twins (non-identical, 50% same DNA) on these outcomes. Or, adoption studies would show children resembling their birth parents, not their adoptive parents. Often, all the resemblance among family members would appear to be accounted for by genetics. Biology rules. After all, your DNA is the ultimate exogenous variable: fixed at conception, it affects you throughout life, but your life doesn't change it.

Or is it?

Take a step back: at individual level, DNA is a fixture of biology (though its effects can certainly interact with the social environment). But how you got your DNA is a function of how Mum met Dad. And that is a social, even a political variable. For an extreme example: many US states banned interracial marriage until the 60s. For a less extreme example, think of"the great sorting out" – the idea that people increasingly live alongside others who are similar to them.

This paper is a first step to thinking about how a social variable – who marries whom – affects the distribution of DNA in society. Social scientists have known for a long time that "like marries like": here we show that, as you'd expect, this means that like DNA marries like DNA, using the Polygenic Score for education that I wrote about last time. (Here's another paper on the same topic for the US.)

There's lots more to do on this topic. As my co-author Abdel Abdellaoui's PhD thesis put it:

behaviour ⇄ genetics.

Tuesday, 23 August 2016

A picture from our forthcoming article in Intelligence

Here is a single bar chart from our forthcoming article in the journal Intelligence. (I've edited it a bit to have nicer colours.)

The bar charts represent individuals' Polygenic Risk Score for educational attainment. That is, each individual gets a score predicting when they will finish education. Don't worry about the details of how this is done: the point is that it combines the effects of many different genes, and it is purely derived from DNA. Give me a cheek swab with your DNA, or your child's and I will calculate your score. Bars 1 to 5 represent the bottom 20% of our sample on this score, then the next 20% and so on up to the top 20% on the right. The bar colours show the proportion of each group that went to university, got A levels (i.e. the age 18 exam in the UK), got GCSEs (the age 16 exam) or left without any qualifications.

This isn't actually what our article is about – our article is on assortative mating between spouses on this score.

The picture just shows that as of 2016, you can predict someone's education fairly well from their DNA. For example, only 1 in 5 of the bottom 20% group go to university, while about half of the top 20% do so.

It used to be that, while genes could significantly predict social outcomes, the amount of variation they actually explained was very small – it was statistically significant but substantively irrelevant. As you can see, that is not really true any longer, and it is likely to get less true as more large-N studies produce more accurate predictions.