Epistasis, and The Mermaid's Tale

Jason Moore spoke about "missing heritability in neuropsychiatric illness" at Dartmouth Grand Rounds yesterday, so of course I was all over that.  It wasn't what I expected - Grand Rounds seldom is.

But via his site, epistasis (gene/gene interaction) I learned about The Mermaid's Tale, a blog by Penn State profs which is much more cautionary about the promise of genetic research, GWAS, and personalised genomic medicine than what I usually read.  You might start with their four-part series Who me?  I don't believe in single gene causation (or do I?) a cleverly-written puncturing of the popular myths - even among the scientifically literate - about genetic research.

I confess I am not entirely persuaded, by those essays and some others.  They quite rightly point out that the popular treatment of genomics is oversimplified and misleading.  Yet I think that is the lot of all complex science in the popular imagination.  Reasserting "But it's more complicated than that" is not so much helpful as a bit defensive.  There is also a good deal of limitations-of-technology rhetoric that seems more aesthetic than simply rational, such as food sustainability assertions founded on the idea that newer technologies are more complicated and fragile.  Yes, and so is modern medicine, communications, and entertainment, all of which depend on electricity and networks.  That's sort of what technology is.

But I quibble.  I agree with a great deal of what they write, and they write well. A simple arithmetic point I had not bothered to consider is worth noting.  The number of possible human beings, possible different genomes, is so huge that even the billions of current examples, were we to analyse them all, are not much different from reading one genome.  What we have as a sample size is analogous to regarding the first inch off the launch pad as a reasonable representation of our trip to Alpha Centauri. (But think of how many thousands of separate nanometers we sampled in that inch! Aren't they a lot?) Therefore, there are meaningful interactions between two, three, or five genes which could definitely "cause" one disease or another, which we will never know about because they will not occur in a living being.  Even if they do occur in one, two, or ten humans, we might not be able to separate that statistically from chance.  Rather like the infinite library in Borges, which contains books both proving and disproving the existence of the library, plus an infinite number of volumes which differ from those only by one or two letter, mere typos.