Bound to Fail

Efforts to bound the heritability of complex traits are silly nonsense

There is an effort, on the behalf of serious researchers, to place a bound on the heritability of complex traits. This effort really makes no sense at all and it is surprising that any serious researcher would participate. Consider intelligence, the real issue here. Considering the immense complexity of the brain, the genetics of intelligence is likely to be at least an order of magnitude more complex than the genetics of height. Really, who knows, it could be several orders of magnitude more complex. More SNPs, but also a greater chance that there are groups of SNPs that make a big difference if they are all present, but little to none if just a few are present. Consider, the brain is a huge signal processing network. If one SNP changes slightly how a signal is sent, its effect may be greatly amplified if some other SNP changes slightly, and in a complementary manner, the way a signal is received.

Our examination of the genetics of the brain is like walking into an immense room, with ornate furnishings and decorations, and we have just a dim flashlight and have only just started to look around. And…we think we can bound the heritability of intellect?? Really??? I think this claim should be met with skepticism. If this group says, “our math proves it” I think it is a good idea to take another look at that math, and how it was derived.

Their technique involves finding the difference between the GWAS trait score versus actual phenotype in general population statistics, vs. the same statistics for siblings. If the general population statistics show a stronger correlation between GWAS score and phenotype, than the sibling statistics, then the GWAS must be inaccurate and factors other than genetics must make up the difference. I see an obvious problem with this line of reasoning.

In the general population, the similarity of GWAS scores is bound to be correlated to the degree of relatedness. Of course, two persons with identical GWAS scores may be completely unrelated. But, they are more likely to be closely related than two persons with very different GWAS scores. They could be siblings or first cousins or perhaps 10th cousins in a hundred different ways (because they are both Ashkenazi).

So, if we consider two pairs of individuals, who have somewhat different GWAS scores, wherein one pair are a priori known to be siblings but the other pair has an unknown degree of relatedness. The pair with the unknown degree of relatedness are somewhat less likely to be closely related, than a pair with more similar GWAS score. Accordingly, all of the immense genome complexity not captured in the GWAS score is more likely to also be different. This would have the effect of increasing the chance that the phenotype is different. But, with the same GWAS score difference the siblings are, a priori, known to be siblings. They are very closely related and accordingly, all the genome complexity not captured in the GWAS score is more likely to be similar, bringing their phenotypes closer together.

QED. Have at it.