A few people have inquired of the PNAS paper On sharing genes with friends. I avoided comment in part because I'm skeptical of the findings. So much behavior genomics just hasn't panned out over the long term, and is probably susceptible to the issues which fuel the "decline effect". Statistical significance is a random variable too. The fundamental issue which I want to emphasize is this:
many behavioral traits are highly heritable, insofar as the correlation between relatives of trait value is in direct proportion to their genetic correlation.
But, just because a trait is heritable does not mean that you can affix the variation to a specific set of genes. That is because the character of genetic architecture varies, and it may be that for many behavioral traits with some biological basis the causal variants which are responsible for the range in trait values are distributed across thousands of genes, and so are of very small effect. Carl Zimmer relayed the depth of skepticism in the scientific community yesterday, and today Dr. Daniel MacArthur reviewed the paper. Here are the top line reactions:
Altshuler’s skeptical view of the paper was fairly widely shared by colleagues I discussed this with yesterday: given what we know about the genetics of complex traits, it seems a priori extremely unlikely to have found two real associations in a study of just six genes, even if those genes have been selected on the basis of biological plausibility. This seems especially unlikely for a behavioural trait: the complete failure of recent large genome-wide association studies to uncover any genetic variant convincingly associated with personality traits suggests that, while these traits are known to be strongly influenced by genetics, those influences are not exerted by common genetic variants of large effect. The buzz amongst the genomics community on Twitter was generally similarly negative, although informed discussion wasn’t helped by the fact that (thanks to the standard ridiculous post-embargo delay from PNAS) the actual paper wasn’t available online until yesterday evening – meaning that there was little besides mainstream media reports for most people to base their judgement on. So, what are we to make of the bold claims in this paper? ... Overall I find myself rather torn here. While David Altshuler is absolutely right that this study wouldn’t meet the criteria for publication in Nature Genetics, and while I’m generally a fierce critic of both candidate gene association studies and behavioural genetics in general, there is more substance to this study than I expected. I’m not saying I’m confident the findings are real – that will require a full, independent replication study looking at exactly the same markers typed in this study – but it’s certainly a result that warrants follow-up. ... So, where does this leave us? Not a great distance from where we started, really. This study is an intriguing observation in support of a broadly plausible hypothesis, and it’s entertaining to consider its implications for association studies, human evolution, and gene-environment interactions. However, until we see the promised large-scale GWAS, it’s best not to spend too much time pondering these implications: let’s save that for if and when we have the evidence needed to confirm that these effects are real, and to gain a better understanding of how common they are in the genome.
More broadly, fixing upon the most recent findings implicating a specific set of candidate gene(s), which only later turn out to be false paths poisons the bigger project of putting in proper perspective the biological component of human behavior. The heritability of many behavioral traits is is a robust finding. This truth often is unjustifiably put into the same category are "candidate gene for trait X." When it comes to genetics and behavior there is the unfortunate reality that the stringency of the evidence needed to persuade is extremely conditional on normative preferences. Much of the media is hungry for genes for every ludicrous behavior (I am pretty sure that a fake press release on a "gene for skate boarding" could result in a series of articles all around the world), and, there is a broad acceptance in the public that behavior is substantially heritable. At the other extreme are those who are skeptical for normative reasons of universal tendencies in human behavior, or variation between sexes and groups. The same people who in other areas accept the provisional findings of "soft" sciences became as demanding as the most rigorous physicist when presented within findings which are not congenial to their preferred outcomes. Because of this consistent structural problem one must always tread with care and prudence when it comes to extraordinary claims.
