On Genetic Heritability

As promised I have a response to this article in the New York Times (I had to spend a couple days marshalling my evidence). I thought I would summarize some evidence about what we know from behavioral genetics so you could understand why I think this article was so wrong. I have tried to classify what we know below by the kind of study that we use to know it. Hopefully this will make things a little bit more clear.

(WARNING: This post is what economists would call a Schumpeterian tome.)

People studying behavioral genetics in humans use three types of studies:

Twin studies: Twin studies compare the concordance -- the appearance of a trait like say liking Britney Spears -- in two individuals. They generally compare the concordance in monozygotic twins (read: identical twins that have identical genes) and dizygotic twins (read: twins that share on average 50% of their genes). Henceforth I will refer to these as Monozygotic Raised Together (MZT) and Dizygotic Raised Together (DZT). This nomenclature indicates one very important fact about these studies: in order to compare genetic heritability these studies rely on pairs of twins that were raised together in a common environment. (Other types of studies discussed later rely on twins raised apart to determine environmental influences.)

How would we use twin concordance to determine heritabiliity. Say that in the groups of twins we survey the MZT concordance (the percentage of monozygotic twins raised together that both have a trait) is 50% for both likely Britney Spears. The DZT concordance (the percentage of dizygotic twins raised together that both have the trait) is 25%. We can use these two numbers plus the knowledge that monozygotic twins share twice as many genes to calculate the heritability. The formula is below.

MZT = monozygotic twin concordance
DZT = dizygotic twin concordance

2*(MZT - DZT) = heritability

2(.5-.25) = 50% heritability for Britney Spears liking

Simple as that. This value equals the rough percentage of the co-occurence of a trait that can be attributed to the common genetics of the offspring rather than other factors.

So what are some levels of heritability in personalities traits such as the ones discussed in the article. An excellent review on the subject by McGue and Borchard, Annual Review of Neuroscience, 1998 has much of the data that I am going to discuss. Pychologists breakdown personality traits into 5 major categories: extraversion, agreeableness, conscientiousness, neuroticism, and openness. They report that the heritablity of these traits is on average about 41% -- leaving a big gap for environment I would say. We could look at similar results for say occupational choice. Heritability estimates (using a scale that classifies occupation into 6 categories) is somewhere in the 35-50% area -- still a lot of environment. OK, what about psychiatric disease? Those at least should have high heritability. Well, in some cases yes and in some case no. Some neurological diseases have 100% heritability, like Huntington's, but actually most psychiatric diseases even have a fair amount of enviromental influence. The chart below indicates some data related to psychiatric diseases (this table is from the review above):

i-1bc428b4dc31fd52e3ea8008e8a36881-table1.jpg

Using the equation we discussed earlier, we see that there is substantial heritability for some diseases like 80% for affective disoder, but not 100%. Even in the extreme cases of psychiatric illness, there is an environmental component in behavior.

Adoption studies: What about this whole environment business though? What can we say about environment?

We can say a fair bit actually, and what we can say relates to another type of study that behavioral geneticists use -- the identical twins separated at birth study. Yes, I know...it sounds like the plot of The Parent Trap -- Lindsey Lohan was so much more normal then -- but this is actually really useful stuff.

i-ea85f03b4fb0c26e3c4778c916f54e5a-98m.jpg

These studies have been done in the US, but they are actually usually done in Scandanavian countries because these countries have elaborate databases for adoption. What we look for again in studies like these is differences in concordance again, but this time differences between monozygotic twins raised together (MZT) and monozygotic twins raised apart (MZA). In this case, we are controlling for common genetics. What we are looking for the difference caused by a common rearing environment -- what psychologists call shared environment -- and other random variables in the environment -- what psychologists call nonshared environment. Below is another table from Bouchard et al. in their 1990 Science study of twins reared apart. They graded twins reared together and twins reared apart on a battery of psychological tasks.

i-8f712bf3770f767b876b230171156b19-00p0025p_l.4.jpg

You want to look at the R values for the MZT and MZA groups, which represent the concordance. For cases where the MZT is greater than the MZA, the difference between the two represents the amount that can be attributed to shared environment.

What we find from studies like this one is that shared environment actually counts for shockingly little of the environmental influences on inheriting a trait -- a huge comfort to parents I have no doubt. In general you are looking in shared environment of about 10% for most traits. There are two noteable exceptions to this trend (this is not in the chart but discussed elsewhere): cognitive ability and juvenile deliquency. Juvenile deliquency seems pretty straightforward, but there is something that you should know about cognitive ability. Cognitive ability -- measured as IQ -- does show about 20-30% shared environment, but most studies of this sort are done in children or young adults. If we measure older twins, we find that as they age the shared environmental factors fall away to genetic factors as they age. This sort of makes sense. Differences in experiences between twins accumulate over time. Thus, it looks like parents effects on their kids cognitive ability relates more to the rate that they attain it rather than the absolute value they attain.

There is another confound I might add just so parents don't get really mad at me. Most adoptive parents fall into a narrow range on the socioeconomic ladder. Thus, they are not handing out adoptive twins to either the very poor or the very rich. This is relevant because it means we may be underestimating the importance of rearing in IQ, primarily because we are screening out all the environments that would be truly deleterious to a child's learning. It would appear that parents are important in providing an environment that is at least conducive to learning, but after that heritability takes over determining the final cognitive ability.

Linkage: This final strategy that behavioral geneticists use to determine whether a gene is heritable is linkage. Linkage involves determining the association with a particular trait in families along side particular molecular variations in the genome. Association between variations and the trait leads scientists to believe that those variations may be located near the genes responsible for that particular trait.

From my experience with the behavioral genetics of schizophrenia (what my lab studies in part), I can tell you that there are two major considerations when it comes to interpreting genetic linkage data when it comes to behavior.

1) 100% of behavioral traits and at least 90% of psychiatric diseases are multigenetic in nature, if they are genetic at all. You are never going to find a single gene whose variation is responsible for anxiety. There are no mood genes. Individual genes code for proteins, not for abstractions. To put this in context, if I were to say that a gene contributes to schizophrenia, we are looking at a gene variant that might appear in perhaps 10% of the affected cases. Many other genes variants could be associated with the disease sometimes together, sometimes not.

2) Behavioral genes always represent an inherited risk rather than an inherited destiny. You have gene variants whose penetrance -- resulting in a psychiatric disorder -- is highly dependent on the environment of the individual. Absent a certain environment, even individuals with high genetic risk will not have the trait or disorder.

Here we get to my example: depression. Caspi et al published a prospective study in Science 2003 where they followed a cohort of children and tried to relate their genotype to their risk of getting depression. For the genotype of this study, they looked at a serotonin transporter gene, 5-HTTLPR, that has a short and a long variant. What they found was that having one or two copies of the short variant of the transporter carried an increased genetic risk of getting depression. However, this genetic risk was only manifest in the presence of stressful life events in adolescence and early adulthood.

Here is a chart from their study:

i-b8b83eebaeabb4fc2eadec26041635a9-se2731691001.jpeg

In the table we see the individuals in the cohort broken down by whether they have two long variants (l/l), one short and one long variant (l/s), and two short variants (s/s) for the serotonin transporter gene. We see that as stressful events increase, the incidents of things like depressive episodes increase with them for all the groups. More importantly, however, the increase in the high risk group (s/s) increases from essentially identical to the low risk group to much higher. These results suggest that their genotype, while it confers a genetic risk, does not confer a genetic destiny to get depression. Depression can be avoided by preventing life stressors.

What have we learned from the three types of studies above? We learned from twin studies that some behavioral traits are heritable, but that heritability varies dramatically from limited inheritance to absolute. We learned from the adoption studies that if you break down environmental factors, you get a lot effect from random or so-called nonshared enivornmental factors. And we learned from linkage studies that even when genes are identified for a behavioral trait, whether or not that trait manifests itself is highly dependent on environment.

All of these pieces of evidence bring the idea of genetic determinism into question. The truth of the matter appears to be that genetics and behavior are vastly more complicated than either genetic determinism or absolute behaviorism can adequately explain. Hence you understand why I was so incensed by this article in the NYTimes. The NYTimes article plays fast and loose with the idea of genetic determinism, downplays evidence that environment matters, and profers conventional wisdom as science. This is unacceptably bad journalism.

I will leave you to look at the article in this light, but the important thing in the future to remember when reading an article like this is that the world is complicated. Scientists aren't lying when they say there are some behavior-related genes, but those genes confer genetic risk rather than destiny -- shooting a whole through this idea that we are no longer responsible for our actions. When we misrepresent this complexity, we set a dangerous and inaccurate rhetorical precedent.

UPDATE: Gene Expression displays rather resoundingly why I am not smart enough to be a geneticist. You should read it though...maybe you can explain it to me.

More like this

I just thought this paper was kind of cool. It reviews the evidence from twin studies that shows that certain regions of the brain show very high levels of genetic heritability. Heritability, as I discussed in an earlier post, is a gross measure of the genetic as opposed to the environmental…
I just read an interesting new paper, Genetic Variation in Political Participation: The decision to vote has puzzled scholars for decades...The results show that a significant proportion of the variation in voting turnout can be accounted for by genes. We also replicate these results with data…
Imagine that there is a trait observed among people that seems to occur more frequently in some families and not others. One might suspect that the trait is inherited genetically. Imagine researchers looking for the genetic underpinning of this trait and at first, not finding it. What might you…
Heritability of the Specific Cognitive Ability of Face Perception: What makes one person socially insightful but mathematically challenged, and another musically gifted yet devoid of a sense of direction? Individual differences in general cognitive ability are thought to be mediated by "generalist…

Are their any twin studies that compare incidents of major psychological disorders?

For instance, most people with schizoaffective disorder have most (or all) of the symptoms of major depression, and anxiety disorders. Given that, you could hypothesize that identical twins, which may have a .48 probability of sharing the same diagnosis, may have a much higher probability of sharing a major psychiatric diagnosis.

Some neurological diseases have 100% heritability, like Huntington's,

a nit, isn't huntington's generally considered a mendelian trait? (autosomal dominant) therefore, technically it isn't 'heritable' in that it isn't a quantitative trait, right?

You make a good point razib. I debated whether to include that as an example, as it is more of what I would call a neurological disorder. I would even go so far as to say that a good definition for "psychiatric disorders" are disorders that have a strong environmental component.

jake, i think there are theoretical reasons why you would have a hard time offering an example of a trait of psychiatric interest that is extremely heritable. the reasoning is as follows

1) traits of great psychiatric interest are common and very pathological

2) pathology ~ less fitness

3) this means that selection should operate against said trait

4) ergo, for the trait to be persistent for such a long time it seems likely that one should weaken the power of selection somewhat. ergo, reduce the % of variation due to additive genetic variance.

Just one little addendum to Razib's point 2). Some psychiatric disorders like autism and schizophrenia can't have had fitness benefits, or break even progeny-wise, in any human society at any time. Thus, pathology ~ fitness.

But some disorders in the DSM are really just ways of codifying the consensus that "it pisses people off when someone behaves this way" -- like ADHD or sociopathy. Ding et al. showed positive selection acting at the DRD4 locus, the 7R allele of which is associated w/ several things, but most strongly w/ ADHD. Harpending & Cochran then expanded on how this might have conferred benefits in some environments. And sociopathy could rise to above the mutation rate as long as they could fool people and as long as they weren't too frequent (which also goes along w/ escaping detection). So it depends on what type of disorder we're talking about.

Ding et al: http://www.pnas.org/cgi/content/abstract/99/1/309
Harpending & Cochran: http://www.pnas.org/cgi/content/extract/99/1/10

This final strategy that behavioral geneticists use to determine whether a gene is heritable is linkage. Linkage involves determining the association with a particular trait in families along side particular molecular variations in the genome. Association between variations and the trait leads scientists to believe that those variations may be located near the genes responsible for that particular trait.

Linkage vs. association

Commenter JP does bring up a good point. I suggested that it was linkage that allows us to find most of these genes, but actually I think I mean disequilibrium mapping -- as this method is much more amenable to finding genes that are much more loosely associated with the diseases they contribute too.

Thank you for the correction.

well, both have sort of been used. The protocol for finding a gene involved in trait in the past was to perform a genome-wide linkage scan, and if there was a region that stood out, then do an association study in that region.

you put your finger on the problem with that--linkage isn't as powerful as association when the penetrance of an allele is low. if only you could do association on a genome-wide scale...

keep your eyes out for genome-wide association studies of psychological disorders in the next few years...

Funny you should mention that. There is a NY Academy of Science mini-conference on just that issue this Thursday for those in the area who are interested.

"Date: Thursday, June 22, 2006
Location: New York Academy of Sciences, 2 East 63rd Street
Frank Middleton, SUNY Upstate Medical University, "When the Chips are Up: High-Throughput Genetic and Genomic Studies of Neuropsychiatric Disease."

Jurg Ott, The Rockefeller University, "Genome-wide Disease Gene Mapping by Association Analysis."

Since I was pointed over here by Lindsay Lohan, might as well comment.... ;-)

Agnostic: Just one little addendum to Razib's point 2). Some psychiatric disorders like autism and schizophrenia can't have had fitness benefits, or break even progeny-wise, in any human society at any time. Thus, pathology ~ fitness.

Unfortunately, your own examples show the fallacy of your claim....

Neither autism nor schizophrenia is a binary condition; they each show a continuum from crippling severity down through "mere handicaps" and onwards to "normal tendencies". And in both cases, the weaker forms can produce people with striking and otherwise-rare talents: Mild schizophrenia can produce creative types, always helpful to have around. Mild autistics have historically shown up as smiths, inventors, animal handlers, and occasionally saints or sages. More recently, they led the Information Revolution, at least since Turing. Both autistics and schizophrenics produce "shamanic" sorts as well, for whatever that's worth.

By David Harmon (not verified) on 08 Feb 2008 #permalink