Genetics in The Atlantic

Share on FacebookShare on Google+Email this to someoneTweet about this on Twitter

A reader points out that David Shenk is blogging genetics & IQ over at The Atlantic. If you have some free time to kill in comment sections, you might be interested. To get a flavor, a post titled The Truth About IQ has a footnote to Stephen Jay Gould’s Mismeasure of Man.



  1. Since you bring up the topic of IQ and genetics, I would like to end gnxp’s deafening silence on the latest news about genome-wide association studies. Since these studies have found no “unequivocally associated” locus for IQ, it seems that a great multiplicity of genes affecting IQ could undermine a significant role for natural selection. Now a study of aortic aneurysms shows that blood samples failed to represent the somatic mutations of diseased and non-diseased tissues. Would this not undermine previous conclusions based on GWAS’s, which rely on blood samples? I also wonder if the presence or absence of such somatic mutations could have functionality.

  2. no IQ GWAS of sufficient size has been published to say with any confidence what the genetic structure of the trait is. however, if you were betting you’d guess that it’ll be like height. that said, there’s been considerable progress on identifying height loci with meta-analyses. 
    given that all complex traits appear to be highly “multigenic” this isn’t a particular problem for IQ. the intersection of genetic architecture and the power of selection would be interesting to discuss.

  3. The hypothesis termed novel by that abstract has been proposed before: 
    Do you know if the mutation in question was found in the majority of aortic or aortic aneurism cells? I’d be about 99% sure the answer would have to be no (for any noncancerous tissue), but I lack access to the paper. Anyway, if the mutant alleles are a small minority in the sample, it won’t affect the results of Sanger sequencing, assuming multiple reads are used. Sanger uses PCR and a minority allele could very very rarely, in theory, wind up being amplified to a higher concentration than the “correct” allele – but if one did say four reads of the sequence in question, this very unlikely event could essentially never happen in two out of the four. I don’t think Sanger is used much now and I don’t know the other methods, but their results are no doubt known to conform to Sanger results. 
    GWAS on IQ have already been done. Thus, to an approximation, we already know there’s “very little or no” variation in IQ due to “common” alleles.  
    If this new study is more powerful, it may be able to change the exact numerical values of “common” and “very little or no” that we can safely use in the sentence above.  
    If memory serves, GWAS of present-day scale are unable to inform us about the effects of alleles whose population frequency is under something like 5%. So rare alleles may be the source of the heritability of g. 
    Or, maybe 10,000 different genes contribute nearly-equal, very tiny amounts to g variation. I can’t recall whether the math allows this possibility to be proven/disproven with any reasonable-sized study, or not. My guess is it cannot.

  4. Whoops, I made a mistake because current GWAS of course don’t use sequencing at all, Sanger or not; they use SNPs. I guess these are typed using microarrays. I don’t know those methods in fine.

  5. “maybe 10,000 different genes contribute nearly-equal, very tiny amounts to g variation.” Is that why it follows a Normal distribution?

  6. I can’t recall whether the math allows this possibility to be proven/disproven with any reasonable-sized study, or not. My guess is it cannot. 
    well, it would seem like you could falsify this. find a larger effect size locus. 
    Is that why it follows a Normal distribution? 
    you can have way fewer loci, and also have something like a geometric distribution, as long as the largest effect size is still small, i think. or at least it “looks normal” in terms of trait value.

  7. to an approximation, we already know there’s “very little or no” variation in IQ due to “common” alleles. 
    Actually, there could certainly be significant variation due to common alleles. 
    What I meant was, from what we know, if any individual common alleles have any effect it is very small. 
    Judging from lol’s words, it seems “very small” was something of an exaggeration. I am certain he knows more about it than me.

  8. Do you know if the mutation in question was found in the majority of aortic or aortic aneurism cells? 
    I shall quote the relevant sections: 
    To establish a reference sequence we examined BAK1 sequence data from a number of databases (up to five) from different Caucasian populations as listed in the NCBI database (Table 1). 

    We collected tissue samples from a total of 31 AAA patients obtained at the time of surgical repair, together with matching blood samples, as well as five samples of normal nondiseased abdominal aortic tissue, that was obtained from a tissue bank. However, when BAK1 cDNAs from actual AAA tissue were sequenced gene alterations were found in three of the amino acids (a.a.) (Table 2), which were exactly the same as the rare SNPs that had been identified in the reference databases (Table 1). In two cases, coding region SNPs resulted in a change in a.a. 42 from arginine to histidine (dbSNP# rs1051911:G>A), and in a.a. 52 from valine to alanine (dbSNP# rs1051912:T>C) respectively (Table 2). In the third case a coding region SNP in a.a. 81 resulted in no change to the amino acid incorporated (dbSNP# 1051913:C>T). This initially suggested that these SNPs could be considered as possible factors in the formation of aneurysms. To confirm these observations, we sequenced BAK1 cDNAs from nondiseased abdominal aortic tissue. Our results somewhat surprisingly showed that the nondiseased tissue contained the same SNPs as the diseased tissue (Table 2). However, when matching blood from the patients was sequenced it was discovered that none of their BAK1 alleles contained any of the reported rare SNPs (Table 2). Thus, the results showed that aortic tissue, whether diseased or not, contained a number of distinct SNPs, that were not present in matching blood samples or in the vast majority (on average >99.4%) of reference sequence DNA (Table 2). However, of interest was that these SNPs had been observed in a very distinct minority of reference sequences (on average

  9. I’m not going to bother signing up and debating it on the Atlantic web site, but I can already see where Shenk in going. To make the argument more clear, I’m replacing “intelligence” with “physical strength”: 
    1) Recent studies have made it clear that a person’s physical strength as an adult is not a fixed, inate quality, but can change dramatically. 
    2) Therefore, men aren’t really stronger than women. 
    Point 1 is clearly true — about strength for sure, and maybe about intelligence as well. Personally it does seem plausible to me that the heritability of IQ has been overestimated based on twin studies. It has never really made sense to me that a person’s intelligence should be entirely unresponsive to training, and what Flynn has to say about this does seem somewhat sensible. Nevertheless, point 2 simply does not follow from point 1, no matter how much social pressure there is to go in that direction.

  10. (continued)

  11. less than 0.06%). 

    In a study of prostate cancer, laser capture microdissection (LCM) was used to dissect prostate tumors into cancerous and noncancerous tissue. Initial results [Alvarado et al., 2005] supported the traditional hypothesis of somatic mutations as the causal agents for carcinogenesis, because genetic alterations in the androgen receptor gene (AR) were found in cancerous tissues, but not in blood. However, in a follow-up study, AR alterations were detected in completely disease-free prostate tissues, remarkably even in prostate tissue from a 1-year-old child [Sircar et al., 2007]. Thus, unlike blood and other tissues, prostate tissue whether diseased or nondiseased and of varying age and maturity, contained variants of the AR gene. 

    The presence of SNPs in nondiseased AA tissue suggests that they may be present prior to appearance of disease and may be a consequence of vascular development and maturation. Their presence might be at least partially responsible for the increased susceptibility of AA tissue to aneurysm formation. A possible alternative explanation of their genetic origin is that within nondiseased AA tissues some BAK1 SNP-containing alleles can exist perhaps as ??minority?? forms, in a similar manner to what we have observed in prostate tissue [Sircar et al., 2007]. That is, they are only initially present in a very few cells within normal aortic tissue and are then selected for when tissue and cellular conditions change. 
    From this I gather that these are genuine differences in genotype between tissue and blood samples. Therefore, GWAS?s could be inaccurate because they assume that blood samples, which I believe is referring to leukocytes, contain DNA equivalent to the target tissue.

  12. Arguably, another mistake that Shenk, Flynn, and others are making is that they don’t take g seriously enough. If g is overwhelming what gives an IQ score its predictive validity, and if it’s impractical to environmentally boost g in adults, then there’s not much more to be said about radically improving people’s abilities. That’s where Flynn’s notion of cognitive training for non-g factors hits up against how seriously you take the evidence for the importance of g across tasks. Maybe you can train people to think more abstractly and yet you’d predict that that won’t do them much good in solving real world problems because it’s g that matters rather than the style of the test questions. 
    For example, consider: 
    The psychometric expert said something that seemed puzzling to me. He said that the General Factor of intelligence completely dominated job performance as a pilot to such an extent that it really wasn’t worthwhile to give multiple intelligences tests of specific piloting skills, such as the one George W. Bush took in 1968 to measure his 3-d visualization skills. 
    My source said that he recommended getting rid of flying-specific tests for admission to pilot-training, but the brass wouldn’t go along with it because they insisted their had to be pilot-specific skills separate from the g Factor.