Mike the Mad Biologist

On Genetic Denialism

Mary Carmichael has a great video (and associated post) about the rise of genetic denialism–ridiculous arguments that ‘genes don’t cause disease.’* Carmichael offers two reasons why the argument is flawed; I’ll offer a third in a bit, but I do want to note one minor point of disagreement with Carmichael.

If you go to roughly the ten minute mark, Carmichael has a summary slide that lists several phenomena that could account for the ‘missing heritability’–the observation that currently identified genomic variation can’t account for much the expected heritability as determined by various studies, such as twin studies or parent-offspring studies (hertability measures how much of the variance in a quantitative** trait (e.g., height) is associated with genetic differences in a population with a certain distribution of genotypes and environments). One thing she lists is gene-environment interactions. That’s actually not entirely correct. Gene-environment interactions (“genotype by environment”), where some genetic variation behaves differently in certain environments, lowers heritability estimates; that is, we underestimate heritability.

What can lead to the overestimation of heritability is genotype-environment correlation. For example, high IQ parents might create a more stimulatory environment for their children (i.e., a ‘high IQ’ environment). This is not a trivial problem, especially since many human studies are not like animal or plant breeding experiments which occur in highly controlled environments. In the comments of one post about this whole issue (which I can’t dig up), someone mentioned how extensive experience with lab models has convinced him to think genotype plays a large role, at which point any evolutionary ecologist–which I was in days of yore–bangs his or her head against the wall***. A little humility about environmental confounding, especially when applied to methods originally designed for controlled breeding experiments would be helpful, along with multiple confirmatory studies using different approaches and cohorts.

Having said that, if heritabilities for some traits are overestimated, this means that the missing heritability problem isn’t as serious as currently thought: genomics has been more successful. Anyway, just a minor point, because what I really want to get to is why I think genetic denialism could become popular.

Carmichael lists two logical fallacies that lead to genetic denialism:

1. It assumes that absence of evidence is evidence of absence, concluding that because common diseases aren’t caused by a handful of genes with strong effects, they aren’t influenced by genes at all. This is clearly a silly position, but it’s a handy one if you are, say, a group that wants to raise the profile of biomedical research into non-genetic factors such as pollution, and if you’re concerned about money being poured into genetic research at your expense. (For the record, I agree that research into environmental factors is important. Maybe the essay’s authors and geneticists should be lobbying together against pending cuts to the NIH? To borrow a malapropism from someone with whom I disagree on almost every other point: shouldn’t we make the pie higher?)

2. It mistakenly portrays science as a monologue, and a dull one at that. First, it lists some of the potential hiding places for heritability that remain in the genome. Then it notes that a few scientists think some of them are less likely candidates than others and that there’s no consensus that one of them will explain everything. This is presented as evidence that almost all of human genetics is in crisis. But the data has just begun to be generated, and of course it’s unlikely that one type of factor will explain everything. Life, in biology as elsewhere, is complicated. (As geneticists freely point out: take it from David Altshuler, Leonid Kruglyak, and a bunch of other people who would know.) We’ve seen this line of “minor disagreement = major crisis” argument before, from creationists who conscripted Stephen Jay Gould and punctuated equilibrium as “evidence” that natural selection wasn’t a widely accepted principle. It was just as absurd — and effective — a rhetorical tactic then as it is now.

I think there’s a third reason, and, while it too is a logical fallacy, it’s far more understandable: a highly heritable trait can be largely environmentally determined. There have been a lot of changes in human health over the last fifty years, such as a real increase in breast cancer rates or the rise in obesity. Even when these changes have a high heritability, such as is potentially the case with obesity, the idea that this is largely response to selection of ‘obese’ genes is absurd. There’s an environmental component. People get this (except for Megan McArdle).

One thing we’ll have to be careful about, and I’m not sure how to do this, is to explain what high heritability means–it doesn’t mean that a trait is ‘genetic’, only that the variation in that trait under a set of certain environmental conditions is largely genetic.

This won’t be easy.

*We can legitimately disagree about the extent of heritability of certain diseases (and other traits). But to argue that there isn’t a genetic component is silly and factually unfounded.

**There are methods to do this with discrete traits too.

***Another issue with genome wide association studies (GWAS) is that the genotype is often much more rigorously specified than the environment. Consequently, “E”, the environmental component becomes a ‘left-over’ term, conflating error and temporal variation with environmental variation. As the Three Toed Sloth noted a while back, we can also run into the same problem were we to overspecify the environment, and then treat the left-over variation as the genetic component.

Comments

  1. #1 harold
    December 28, 2010

    The only time that genetic influences absolutely predominate is when they are quite extreme, for example chromosomal imbalances or single mutation diseases. Likewise, when environmental influences are extreme they can massively impact on a large segment of the exposed population, with little regard for individual genetics, except possibly very specific resistance – e.g. the “black death” (probably bubonic plague) of the fourteenth century, ingestion of radium in watch factories, and so on.

    The rest of the time, health is affected by both environmental and genetic factors acting in complex interaction. And arguably by what humans perceive as random chance as well.

    If there is a wholesale tendency to underestimate the role of genetics in disease, then it may be a reaction to a similar wholesale tendency to overstate the role of genetics in almost everything. Also, while groups that criticize modern industrial agriculture and conjecture about environmental carcinogens may tend to exaggerate and over-interpret, their basic idea is not entirely unreasonable.

    The fact is that genetically identical or similar human beings get different diseases in different environments. Seventeenth century Europeans frequently died of infectious disease at young ages; their descendants in rich countries don’t.

    Cardiovascular disease was common in seventeenth century Europe, but it increased, even when survival to mature age is taken into account, during the first half of the
    twentieth century. Cigarette smoking, sedentary lifestyle, and consumption of industrial “trans” fats are pretty unequivocal risk factors; some types of natural saturated animal fat may play a role as well, at least in some people. On the other hand, cardiovascular disease risk shows high heritability as well, in some cases for reasons that can be shown to be single gene related.

    Some cancers are well known to be associated with certain environmental insults, such as smoking, exposure to certain chemical or radioactive agents, etc. Some environmental factors increase the risk of some cancers while decreasing the risk of others, sunlight being a very strong example of this.

    In other words, looking for genetic predisposition does not (or should not) mean denial of environmental factors, and vice versa. Finding one does not rule out the other.

  2. #2 theshortearedowl
    December 28, 2010

    Genetic denialism… now I’ve heard it all.

    Another thing Carmichael doesn’t mention (at least in the blogpost, I haven’t seen the video) is that our main current method for tying genotype to phenotype – association mapping – is biased to finding single, large-effect genes. There are likely to be many genetic effects that are simply missed because of this.

  3. #3 hjk
    December 28, 2010

    Bravo! I find this to be the most cogent treatment of the nature v. nurture argument I have seen in the 40 years of my professional life (retired school psychologist).

  4. #4 Rich and Co.
    December 29, 2010

    Here is the crux issue we see — any explanation/metaphor/frame that does not put immediate conscious control in the drivers seat is doomed. (Effectively) no one will accept that. The notion that anything (genes?) is determinate challenges pretty much every belief system — especially in the US.

    Free will must rule.

    Current science, especially brain science, is proposing an Copernican revolution but even worser. Our (conscious)minds are not the center of anything. We are biological machines.

    The environment frame just = “we can change (educate/regulate/nudge/wish/pray/etc.) it.” The most compelling fairytale of all. If only.

    Most people still don’t believe the earth is not the center of the universe. Of course it is, dummy!!??

  5. #5 Keith Grimaldi
    January 13, 2011

    You’re absolutely right -. heritability is a very difficult thing to get hold of, it’s often used incorrectly, often confused and often misunderstood (I should know…)

    In the end we want to use genotype to predict future phenotype and GWAS so far is not finding the info we need – is it because of rare variants? Maybe. But it certainly will also be something to do with the lack of precise “E” measurement you mention.

    This recent paper in yeast gives a glimpse of the problem 8http://bit.ly/hbUNpR): Two strains of yeast and 4 SNPs – the effects of the SNPs were NOT predictable unless both the genetic background and the environment were known. There were also SNP-SNP interactions, but even these depended on the environment and genetic background. The complexity in this relatively simple system is not great news for predicting the conditions of disease risk in healthy humans

  6. #6 Mary Carmichael
    January 19, 2011

    Mike,

    Thank you for this! And sorry it took me so long to see it.

    You’re right, I should have been much more clear about gene-environment interaction vs gene-environment correlation. I’d use the excuse that I only had about three sentences to explain that point, but you’ve done it very well in three sentences here, so all I can say is “bravo.”

    Also agree that a big part of the overall confusion is in how regular people define “heritability” compared to how geneticists define it. I tried really hard in the video to make sure I used the proper geneticist’s definition (“it’s the variation, stupid”), but explaining exactly what that means without resorting to equations is challenging. I’ve been trying to think of a way to write a “what the heck is heritability” article since ASHG. Any ideas would be more than welcome!

    Mary

    PS: @theshortearedowl, the video does contain quite a bit of discussion about the limitations of SNP chips and there was also a subsequent untaped discussion that went into more depth. I figured most people reading the blog post (which was more specifically aimed at biology types than the video was) were already all too familiar with the GWAS/missing heritability debate….