Relatedness disequilibrium regression explained

Whether resemblance between relatives is due to genes (nature) or environment (including nurture) has generated much controversy, especially for traits like education and intelligence. The modern scientific effort to disentangle nature and nurture can be traced, in large part, back to Charles Darwin’s cousin, Francis Galton. Prompted by Darwin’s theory of evolution by natural selection, Galton searched for evidence that hereditary factors played an important role in shaping human abilities, especially mental abilities and ‘genius’. He examined whether men that had reached ’eminence’ in some field (literature, science, law, among others, including wrestling!) were more likely than the average to have male relatives who also achieved eminence. One wonders whether his method was inspired by his own illustrious pedigree, the Darwin-Wedgwood family, a family that included Erasmus Darwin, an enlightenment thinker; Josiah Wedgwood, a key figure in the industrialisation of pottery; Charles Darwin; and Francis Galton himself.

Francis_Galton_1850s
Francis Galton in ‘middle life’. From Karl Pearson’s The Life, Letters, and Labors of Francis Galton. 

Galton’s analyses in his 1869 book Hereditary Genius showed that eminent men were much more likely than average to have an eminent male relative. While Galton interpreted his results as implying a strong role for hereditary factors, they could also be explained by environmental factors shared between relatives: growing up in a distinguished family undoubtedly has an effect on one’s development and ambitions.

The underlying logic of Galton’s analysis is that greater resemblance of relatives over non-relatives is evidence for nature over nurture. Since Galton, more sophisticated methods have been developed for disentangling nature and nurture, but almost all share a similar rationale.

Galton himself saw promise in the study of twins, which has since developed into a sophisticated scientific field. Twin studies have been held up as the ‘gold standard’ for measurement of heritability, the fraction of trait variation in a population that is due to genetic inheritance. The most common type of twin study compares the similarity of identical (monozygotic) to non-identical (dizygotic) twins. The studies assume that any greater similarity of identical twins over non-identical twins must be due to the effects of genes. If identical twins experience more similar environments than non-identical twins, then this could lead to overestimation of heritability. While measures of the environment have typically shown more similar environments for identical twins compared to non-identical twins, the exact impact this has had on twin heritability estimates is hard to quantify due to the difficulty of measuring all of the relevant environmental variables (see Felson 2014).

In the past decade or so, an industrial revolution has occurred in genetics of similar impact to the industrial revolution Galton’s ancestor, Josiah Wedgwood, instigated in pottery. New technologies have enabled scientists to measure hundreds of thousands of genetic variants in hundreds of thousands of people, making new methods for estimating heritability possible.

One such method, the ‘Kinship’ method, looks at how trait similarity changes with genetic relatedness for all pairs of individuals from a population (see Zaitlen et al., 2013). If, for example, more related people tend to be much more similar in height than less related people, the Kinship method would say that the heritability of height is high. The problem is that more related people tend to have more similar environments than less related people. Therefore, what the Kinship method ascribes to genes could just as easily be environment.

Since 2010, Peter Visscher and Jian Yang at the University of Queensland have pioneered a method called GREML (see the website for their software, GCTA). GREML is similar to the Kinship method, but it restricts the analysis to distantly related individuals. By restricting the analysis to distantly related individuals, the method avoids being misled by certain kinds of environmental effects shared between close relatives, such as siblings. However, there is one important kind of environmental effect that GREML is vulnerable to: indirect genetic effects from relatives or ‘genetic nurture’.

Scientists at deCODE Genetics, including myself, my supervisor Augustine Kong, and deCODE Genetics CEO Kari Stefansson, demonstrated that genetic variants in parents influence the educational attainment of their children, whether they are passed onto their children or not. This work was published in Science earlier this year (link) and featured in the New York Times. The genetic variants in the parents exert their influence on their offspring through the environment the parents provide for their offspring’s education. This provides a twist to the nature-vs-nurture debate: nurture is itself partly genetic. To understand the implications of this, consider that you inherited a piece of DNA that helped to raise your educational level, then it probably also raised the educational level of the parent from whom you inherited it, giving it two effects on you: a direct effect from inheriting the variant (part of the heritability) and an indirect effect due to the environment created for you by your parent’s educational level. GREML methods do not use genetic information on parents, so are unable to separate direct effects due to inheritance of a genetic variant from indirect effects that act through the family environment. This can lead to overestimation of heritability for traits in which genetic nurture plays a role.

In 2006, Peter Visscher developed a different method for estimating heritability, Sib-Regression. Siblings vary in the amount of DNA they share inheritance of from their parents. To explain where the variation comes from, recall that there are two copies of each piece of DNA in each parent. Whether a sibling inherits one or the other copy of a piece of DNA from a parent is like the outcome of a fair coin toss. The coin toss represents the outcome of the random shuffling of genetic material in parents during production of sperm or eggs. Depending on the outcome of the random shuffling of DNA in parents, siblings could inherit the same piece of DNA from both parents, the same piece from only one parent, or different pieces from both parents. Sib-Regression looks at how much more similar the traits of more related siblings are than less related siblings. The variation in genetic similarity between siblings is generated by random shuffling of genes, which is independent of the environment. This implies that Sib-Regression can estimate heritability in a way that is effectively uncontaminated by environmental effects. While Sib-Regression is an elegant idea for estimation of heritability, in practice it requires genetic information on many more siblings than we currently have data on.

In a paper published today in Nature Genetics by myself, Augustine Kong, Kari Stefansson, and other scientists at deCODE Genetics, we introduce a novel method for estimation of heritability, Relatedness Disequilibrium Regression (RDR). Like Sib-Regression, RDR takes advantage of the variation in relatedness due to random shuffling of genes during production of sperm and eggs in parents, which we term ‘relatedness disequilibrium’. The difference is, by using genetic information on parents, RDR can use the relatedness disequilibrium from all pairs of individuals in a sample, not just siblings. By using a large sample of Icelanders, we were able to estimate heritability precisely for a variety of human traits.

The results we obtained imply that the heritability of some human traits may be substantially lower than currently believed, and that other methods may have overestimated heritability. For example, the Kinship method estimates the heritability of height to be 78% in the Icelandic data, in agreement with twin studies, which put the heritability at around 80%. RDR, in contrast, estimates the heritability of height to be only 55%.

For educational attainment, there is mounting evidence for a substantial contribution of genetic nurture. Consistent results have been observed in Iceland and in populations analysed by the educational attainment genetic association consortium (see Lee et al., 2018). As described above, GREML methods overestimate heritability in the presence of genetic nurture. We found evidence that GREML methods are likely to have overestimated heritability by around 70%, consistent with the estimated magnitude of genetic nurturing effects. We found evidence that the Kinship method has greatly overstated the heritability of educational attainment, suggesting that a recent study employing a variant of the Kinship method may also have overstated the heritability of educational attainment (see Hill et al., 2018).

While the RDR results obtained in Iceland need replication in other populations before strong conclusions can be reached, they suggest that some human traits could be substantially less heritable than currently believed on the basis of twin studies and other methods. Why does this matter? For one, heritability is the ruler against which genetic association studies and genetic predictors of human traits are measured. We need to make sure we are using the right ruler when measuring these things. If genetic inheritance is less important but genetic nurture is more important than currently believed, this has important implications for understanding the causes of intergenerational inequalities. Furthermore, if human traits are less heritable, this implies that greater importance should be given to understanding causes of human trait variation that act through the environment.

 

 

Advertisements

2 thoughts on “Relatedness disequilibrium regression explained

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s