|The calculated "heritability" for having two hands is essentially zero.|
Image: Wikimedia Commons/Leon Brocard
Tuesday, May 16, 2017
WHAT IS HERITABILITY, REALLY?
In the May issue of Scientific American, I critique the latest efforts to find genetic variants that provide elevated risk for schizophrenia, the most debilitating of all mental illnesses. The story is behind a paywall, although I hope that readers will be able to get access to it either through personal or institutional subscriptions. I did, however, provide a synopsis of the main points in an interview with John Batchelor, which is available on his podcast feed.
The article included a sidebar about the heritability of schizophrenia, and some common misunderstandings about what heritability actually is. Lack of space made it impossible to go into much detail, but below I provide an expanded version of the text which includes some additional details. I hope readers will find it useful.
* * *
Researchers have been searching for schizophrenia related genes for at least 50 years. What makes them think they will find them? The rationale is spelled out in the introduction to nearly every scientific paper on schizophrenia genetics: The disorder has a high “heritability.” This term is often interpreted—by many researchers and the general public alike—as a measure of the relative role played by genes. Heritability is usually expressed as a percentage between zero and 100%.
Scientists have estimated the heritability of schizophrenia using several approaches, including studies of twins, both identical and fraternal. One oft cited study dates to 2003, when a research team reported a “meta-analysis” of 12 previous twin studies. (In a meta-analysis, the data from earlier studies are pooled to increase statistical power.) The team concluded that schizophrenia had a heritability of 81%.
However, many researchers argue that heritability estimates for schizophrenia and other so-called complex human traits (ranging from disease susceptibility to how tall a person is) can be very misleading. One major debate is over key assumptions used to simplify the method. One assumption is that genes and the environment do not interact but have only an additive effect; another is that genes act independently rather than in concert. Still another, called the equal environment assumption (EEA), considers both identical and fraternal twins to be subject to the same environmental influences. Thus if identical twins are more similar than fraternals for a particular trait, that greater similarity must be entirely due to genes. But critics argue that the EEA is violated in a number of ways, including the greater likelihood that identical twins will be treated the same by their parents while they are growing up.
“These basic assumptions are wrong,” says Roar Fosse, a neuroscientist at the Vestre Viken Hospital Trust in Norway, who led a critical assessment of the EEA published in 2015. But twin researchers have mounted a vigorous defense of the approach, countering that even if the EEA and other assumptions are oversimplifications, the methodology is basically sound. “I don’t think it’s likely that current heritability numbers are substantially overestimated,” says Kenneth Kendler, a psychiatrist at Virginia Commonwealth University’s School of Medicine.
But some researchers have an even more profound critique of heritability. They argue that it is not truly an indication of the relative role of genes and environment. The actual definition of the term, they point out, is much more technical: Heritability measures how much the variation of a trait in a particular population—whether height, IQ, or being diagnosed with schizophrenia--is due to genetic variation among the individuals in that population. “Heritability and genetic cause are not the same,” says Mary-Claire King, a geneticist at the University of Seattle. Peter Visscher, a geneticist at the University of Queensland in Australia, agrees. “It is a misconception that a high heritability implies genetic determination. Human height has a heritability of 80%, and yet environmental factors such as childhood nutrition and healthcare can have a big effect on adult height.”
As an example of how misleading heritability estimates can be, Eric Turkheimer, a behavioral geneticist at the University of Virginia, points to the human trait of having two arms. Nearly everyone in a given population has two of them, and there is normally no difference in the number of arms between identical twins—who share 100% of their genes—and fraternal twins, who are assumed to share 50% of their genes. Thus when heritability for arm number is calculated, it comes out to zero. And yet we know that having two arms is almost entirely genetically determined.
Figuring out what heritability for schizophrenia actually means is key, researchers say, because even the most high-powered genetic studies have only identified about a third of the predicted genetic component. A similar predicament faces researchers working on other complex diseases, including diabetes and Crohn’s disease, where an even higher percentage of the heritability remains unaccounted for. Will this so-called “missing heritability” eventually show up in more sophisticated studies—or will it turn out that genes are not playing as big a role as heritability estimates have long predicted? The jury is still out.