Monday, April 9, 2012

SuperSize me! Nothing in American can be small (except genetic risks?)

In evolutionary biology, perhaps especially human evolution and anthropology, and biomedical genetics the current working, we mean 'model' of strong, rapid, definitive natural selection as 'the' mechanism by which traits we see today got here.  Since adaptation only works through what is inherited (environmental effects, so to speak, die with the individual), the same kind of simple-cause deterministic thinking has been applied to the genetic control of current traits.

There are all sorts of reasons to expect, or hope, that cause and effect will be simple.  Single-gene causation of adaptation means we can find 'the' gene that explains why you vote or mate as you do, have a particular  disease or physical trait, and so on.  Pharma doesn't want to invest in profit-less rare traits, or complex traits for which a single med will only help a small fraction of patients.  And, of course, simplicity lends itself to melodrama and hence to the visual and even the print news.

But what we see are a multiplicity of individually small effects, as last week's papers on autism (the subject of our post on Friday) show yet again.  This is disappointing, but why is nature that way?  There are several reasons to believe that the apparent complexity is, in fact, the truth.

This should surprise no one.  For example, mutations conferring simple strong effects on disease-susceptibility will be quickly eliminated by natural selection.  Genes fundamental to many other genes because of interactions, may be specifically vulnerable to such mutations--so we may not find many risk alleles in  those genes. 

If many genes contribute to a trait, their individual effects almost necessarily will be even smaller.  This clearly is the case for the kinds of traits that are the main targets of GWAS and similar approaches.

Most genes that confer high risk would be eliminated by selection unless, as some argue, recent environments make them harmful (e.g., causing diabetes or cancer), whereas they weren't harmful before.  If their effects were slight or of late onset, they would not impair reproductive success, and would stay around in the population.  This doesn't seem to be the case.  In most GWAS'ed traits, risk has risen rapidly and greatly during the past century.  Yet the evidence is not that a few genes with major response to these environmental changes are responsible for the disease: indeed, the GWAS problem is precisely that this is not what we find!

Note also that traits not present at birth, meaning most GWAS'ed traits, take decades to manifest themselves.  The risk difference between variants at the 'risk' genes is usually very small, meaning that they change the risk at any given age by trivial amounts.  We may not want to get such diseases, but from a biological point of view they are really miniscule effects.  This also easily and non-suprisingly accounts for the findings of the recent paper of  low concordance of age and cause of death relative to genotypes in identical twins.

The very same arguments apply to the ability of natural selection to detect these differences, and that in turn clearly explains why it is so difficult to find 'signatures' of natural selection in genomic data, and why again in turn most selective arguments that refer to specific genes are without strong support beyond neat stories one tells about them (as we see in the news almost daily, and report here on MT).

When a gene has a true, but tiny, affect on risk (or on evolutionary fitness), there are so many competing causes of death or disease, or bad luck, that the odds on that gene's effect actually being manifest (as disease, or fitness) are simply very very small.

These are not complicated ideas to understand!  They are not our own private theory.  They're plainly visible in the mountain of facts we already have available to us (without huge, costly biobanks and promises of personalized medicine or strong adaptive arguments).

Traits like disease or adaptation may be major--nobody wants cancer, but in trying to find 'the' gene or few genes that are responsible, we're making mountains out of biological molehills.


John R. Vokey said...

Not sure whether you have seen this:

but the results are consistent with your principle argument.

Anne Buchanan said...

Thanks, yeah, we blogged about that story on April 2 ("Ho hum, you're SO close to average"). The message is -- we think -- that genetics money should be spent on real genetic issues, and the hype about personalized genomics and what your genome sequence can tell you about your risk needs to stop.