Underestimating heritability


Jason Collins


August 25, 2011

It’s not normally a good sign when an attempt to skewer measurement of heritability opens with a link between genetics and eugenics via Francis Galton, and Brian Palmer’s critique of twin studies is no exception (HT Razib at Gene Expression).

Twin studies are one of the primary methods to estimate the heritability of a trait. As monozygotic (identical) twins are more genetically similar than dizygotic (fraternal) twins, the similarity between identical twins and fraternal twins can be used to infer heritability. Palmer argues that this technique has a flaw:

That identical twins do not, in fact, have identical DNA has been known for some time. The most well-studied difference between monozygotic twins derives from a genetic phenomenon known as copy number variations. Certain, lengthy strands of nucleotides appear more than once in the genome, and the frequency of these repetitions can vary from one twin to another. By some estimates, copy number variations compose nearly 30 percent of a person’s genetic code.

These repeats matter. More than 40 percent of the known copy number variations involve genes that affect human development, and there are strong indications they explain observed differences between monozygotic twins. For example, it’s often the case that one identical twin will end up victimized by a genetically based disease like Parkinson’s while the other does not. This is probably the result of variations in the number of copies of a certain piece of DNA. Copy number variations are also thought to play a role in autism spectrum disorder, schizophrenia, and ADHD, all of which can appear in only one member of a monozygotic twin pair (PDF).

This point is starting to be raised more regularly in attacks on twin studies, but  it is not particularly damaging. In fact, to the extent that identical twins are less genetically similar than assumed, it is possible that heritability is being underestimated. If the similarity in traits occurs despite less than identical genetic similarity, this suggests that we should attribute more of the similarity to genes. Of course, if fraternal twins are also less similar than assumed due to these same effects, we are back where we started. However, we are no worse off than our starting point. I’ll start to worry when results from other techniques start to contradict twin studies, but so far, estimates of heritability from genomics are pointing in the right direction.