Friday, September 12, 2014

So...it's not genetic after all! (but who's listening?)

Is it time to predict the gradual out-cycling of a focus on genetic causation and a return of environmental causation, in our mainstream scientific dialog (and funding-pot)?  Such a recycling is bound to happen--even if, say, genetics' claims were correct.  Why?  Because the new generation has to have something to show they're smarter than their elders, and because the abuse of genetic determinism by society is a nearly inevitable consequence of the fervid love-affair we're now having with genomics and its glittering technology.  But maybe there's another reason:  maybe genetics really has been oversold!  Is it possible?

Bees and societal (in)determination
Honey bee harvesting is a social phenomenon and experiments by various authors have found that only a fraction (in some studies, 20%) of the workers actually do most of the work.  But a recent controlled study reported in the journal Animal Behavior by Tenczar et al (vol. 95, pp41-48, 2014, but paywalled) found that if those 'busy-bees' are removed, others step in to fill the work gap.  The gist of the evidence seems to be that among the gatherer work force (and presumably other castes as well, though that's not reported), there is a spectrum of contribution and it's condition or context-dependent.  As the paper says:
These bees resembled elite workersreported in a number of other species. However, our results also show that honeybee foraging activity level is flexibly adjusted during a bee's lifetime, suggesting that in honeybees, elitism does not involve a distinct subcaste of foragers but rather stems from an extreme of a range of individual activity levels that are continuously adjusted and may be influenced by environmental cues.  . . . these results support the view that individual workers continuously adjust their activity level to ensure that the colony's nutritional needs are being adequately and efficiently met, and that the net activity of the whole foraging population is likely to be one of the factors that influences this decision. 
The authors discuss the fact that these patterns have not been studied, with varying levels of rigor, in many species of social insects.  While it is not clear that genetic differences are never partly responsible, the evidence is that social roles are not rigidly pre-programmed.  This study was presented by the NYTimes with a captivating video from the authors, but while that was nice and led us to the research story itself, the Times characterized this as a system allowing upward social mobility.  That's a bit pandering to middle-class readership, and didn't really critique this work in the context of today's prevailing genetic-deterministic viewpoint. However, the idea of context-dependent roles, based on the needs and opportunities in society at large, is noteworthy and of course is something that also happens in humans.

Honeybee; Wikimedia Commons

This of course raises the question of how the bees perceive the needs or different roles, or if the role pattern is a spectrum of activity of each bee, then how does it know when and what to do.  This would relate to the bees' brains' ability to digest quite complex information and make decisions, something very interesting to try to understand, and something we wrote about here not long ago.

Intelligence
A new paper in PNAS reports the results of a large study of the genetics of IQ.  Essentially, they found three genes with very small effect and unknown functional association with cognition.  Indeed, one of the genes may not even be a gene. To sort this all out, of course, they say they would need a sample of a million people.  One of the authors faced with this mountain of chaff is quoted this way in the story:
Benjamin says that he and his colleagues knew from the outset that their efforts might come up empty handed. But the discovery that traits such as intelligence are influenced by many genes, each having a very small effect, should help to guide future studies and also temper expectations of what they will deliver. “We haven’t found nothing,” he says.
Nice try!  But the truth is that that is just what they have found: nothing.  Or, at least, nothing new, that is, no thing.  We knew very well that this was the most likely sort of finding.  We have countless precedents, including the results of countless earlier searches for genes for intelligence (and, for that matter, similar findings for most psychological/behavioral traits).  Like other traits from normal ones like stature and IQ, to body weight and major diseases of all sorts, we find polygenic control--countless contributing genetic factors with individually minimal effect. This even though usually the heritability of the trait is substantial, meaning that variation in genes together accounts for a non-trivial fraction of the overall variation in the trait (the environment and other factors contribute the rest, usually around 60-70%).  

But heritability is a persistently subtle and misunderstood (or ignored) measure. Even with nontrivial overall heritability, the aggregate nature of the measure means we cannot say in any given individual whether his/her IQ is based on this or that particular genes, or is some specifiable percent due to genes (that is itself difficult to make sense of when referring to an individual).  And heritability is often measured after taking out, or controlling for the major real causal factors, such as age and sex.  Arguing for a sample for a million, if allowed and funded, is a huge fool's errand and a corrupt way to spend money (because it's mainly to keep professors off the street of unemployment).

Yet the issues in these cases are subtle, because we also know of many different individual genes that, when seriously mutated, cause direct, major, usually congenital damage to traits like intelligence.  Yet few if any of these genes show up in these mega-mapping studies.  It is this sort of landscape of elusive complexity that we need to address, rather than just building expensive Big Data resources that will largely be obsolete before the DNA sequence is even analyzed, based on the daydream that we are not, knowingly, chasing rainbows.

The primary question one thinks to ask is whether 'intelligence' is a biologically meaningful trait.  If not, even if it can be measured and be affected by genes, it isn't really an 'it' and one can't be surprised that no strong genetic influences are found even if the measure is stable and heritable.  Asking about the genetic basis of intelligence under such circumstances is not asking a well-posed question.

Baby stories
The other day we posted about the recent Science issue on non-genetic influences on parenting,  environmental effects on traits and how long-term and subtle they can be, and how they are not Genetic in the sense of the G-rush we are currently experiencing.  The stories are many and diverse and tell the same tale.  

Here the fascinating question is how the various environmental factors could influence a fetus in factor-specific manners that even relate to the factor itself (e.g., maternal diet affecting the future baby's obesity level, or the effect of the mother eating garlic or being exposed to odors on taste preference or specific odor-related behavior in the child).  To answer such questions we have to know more than just about a gene or two.

So, why aren't these findings grabbing headlines?
The bee story made the front-page of the NYTimes, but mainly because of the video and not because it is a counter to the strong genomic hard-wiring ethos so often promoted by scientists these days.  Likewise, the baby influences made the cover of Science, but we didn't see a Hot-News blare announcing that genetics isn't, after all, everything.  And of course the IQ story didn't make that clear either, given that the author said he wanted studies of a million to find the real genetic causes of IQ.  And, determinists say this isn't going to change their mind about the genetics of intelligence, because it's definitely genetic.  

Will we, or when will we, see people begin to back off their claims of strong genetic determinism, and begin addressing the really hard questions concerning how complex genomes interact with complex environments to produce what we are clearly observing?  In my opinion, these questions cannot be addressed from a genetic, or from an environmental, or from a simple Gene + Environment point of view.

7 comments:

Anonymous said...

Give a bunch of psychologists tons of NIH money, tell them to find out a way to use human genome ('because we spent a billion dollars sequencing it and must show some use') and this is what they end up doing !!

In the meanwhile, Dan Graur is out of money for telling the truth.

http://judgestarling.tumblr.com/post/97259387131/eric-green-funding-authorship-integrity-and-an

Ken Weiss said...

Reply to previous comment:
Big Science and a reliance on hypothesis-free (as it's somewhat wrongly called) science has its good sides and bad. The good is it gives us a look at data for the insightful to think about and then design actual science. The bad side is that it entrenches a funding siphon and makes the comfy substitution of technology for thought. We are not good at balancing the value we get for the value we lose.

Dan Graur speaks his mind. Somebody has to do that, if we're to have any tempering of science fads at all.

Unknown said...

Hi Ken,
Could you explain a little more what you mean by Big Science being somewhat wrongly called hypothesis-free?

Ken Weiss said...
This comment has been removed by the author.
Ken Weiss said...

Reply to Dan
It is often said, even by the Genome Institute not so long ago, that they don't believe in hypothesis-based science. That is an excuse to go fishing (in expensive ponds), that is, to justify Big Data without having a focus or a well-posed question.

It was somewhat disingenuous in my view. There is in fact a strong hypothesis: that genetic variation is a contributor to the trait's variation, but we just don't know what genes are involved. It's a strong generic assumption but a specifically weak one, based on the article of faith that, to caricature it just a bit, genetic variation is importantly involved in everything. One is not even obliged to provide any direct, relevant evidence for that hypothesis before launching huge mapping studies, for example. Too much to go into here, but we've talked about that in the past on several occasions.

There are certainly times when this kind of fishing may be called for, but most of those times have passed, because we know enough now not to keep throwing resources into minnow ponds, with low payoff. What has been found so far was long-ago predictable (and predicted).

Now, the faith that genetics is the thing is still clinging on despite increasingly frequent reports of non-genetic (even very early-life) environmental causal influences. Again, as with genomics, it is presently largely a grab-bag in that often just some instances are noted in claims that this is widespread.

Maybe it's a turn of the nature-nurture cycle, maybe just a matter of funding strategy--but also, it reflects new specific findings that are showing us that these non-genomic causal effects are real.

Again, the issue is whether one has to have, or should have to have, a specific hypothesis--or causal theory of some sort-- to test, or whether we should just disavow any such hypothesis (except that environments are important) and again go fishing in a different and even more problematic environmental minnow pond.

This, at least, is a cursory summary of what I meant.

Unknown said...

Hi Ken, thanks for the explanation. I agree with you, I just hadn't considered the inherent assumption that there is something to find.

Ken Weiss said...

Reply to Dan:
Of course genes are involved in everything so it is a safe kind of assumption. But it is more than that, really, because it assumes that genetic variation is a useful, meaningful contributor to a trait's variation. Most traits do have nonzero heritability, but the tacit assumption is predictive power from genotype to phenotype.

For example, carbon atoms are needed for life, but we don't enumerate each person's carbon atoms to predict their diabetes or stature because the contribution is a causally empty one relative to the objective.