Tuesday, October 26, 2010

Postulating causation

A prominent geneticist was on campus the other day to give a talk about what genomewide association studies (GWAS) can tell us about biology.  While he is more favorable about this method than we are, we were pleased to hear him say that GWAS alone are not enough, and that generally this method can, at best, be expected to identify a gene or genes that act in concert with other genes to produce a phenotype.  That is, polygenes -- a polygenic trait is affected by many genes and its variation is due to the joint contribution of variation in many genes.

But we aren't here to debate the merits of GWAS today -- we've done that often enough before.  We want instead to talk about the problem of demonstrating causation.  The speaker said that GWAS can be used to zero in on candidate genes, but once candidates are identified, the researcher must then demonstrate the function of these genes and their relevance to the trait or disease of interest.  And to do that, he said, we need a set of 'molecular Koch's postulates' applicable to genetics.

Robert Koch was a German physician and bacteriologist, and winner of the Nobel Prize in Physiology or Medicine in 1906 for his work on diseases such as anthrax, cholera, tuberculosis and rinderpest.  In 1890 he published a set of criteria that he believed should be used to establish the cause of infectious diseases.   These are the 'Koch Postulates', still in use today:
  • The bacteria must be present in every case of the disease.
  • The bacteria must be isolated from the host with the disease and grown in pure culture.
  • The specific disease must be reproduced when a pure culture of the bacteria is inoculated into a healthy susceptible host.
  • The bacteria must be recoverable from the experimentally infected host.
The corollary postulates for demonstrating molecular causation are these:
  • the wildtype version of a gene must lead to the wildtype phenotype
  • the mutant version must lead to the mutant phenotype
  • mutant + wildtype genes must lead to the wildtype phenotype (that is, the wt must rescue the phenotype)
  • mutant + mutant genes must lead to mutant phenotype (that is, the human mutant allele must lead to the mutant phenotype when inserted into a mutant mouse)
And, demonstration of causation includes study of the biochemical effects of a mutation, as well as the cellular, tissue and organismal phenotypes.

Insistence on taking GWAS beyond the identification of candidate genes is important, and is in large part a response to the many criticisms of the method.  It doesn't mean that the method will now identify single genes with large effects, as advertised, of course, but it does mean that the onus is on the researcher to back up his or her claim of genetic causation. So we applaud the idea of molecular Koch postulates.  

There are well-known problems with the original Koch Postulates, however, and one should take care when using them as a model for demonstrating causation.  Many microbes can't be grown in the lab (the bacterium that causes leprosy is one example), and there are no animal models for a number of human infectious diseases.  And indeed, with respect to postulates for demonstrating molecular causation, many transgenic mouse experiments have shown 'no phenotype' when the mutant allele is introduced into the host, and there is often great variation in phenotype among transgenic litter mates. 

So, the call to researchers to demonstrate causation is an excellent one, but all the usual caveats about how difficult that can be still apply.  


Holly Dunsworth said...

Oh so helpful. Thanks Anne!

James Goetz said...

"The specific disease must be reproduced when a pure culture of the bacteria is inoculated into a healthy susceptible host."

I agree with all of this in theory, but the above might cross ethical lines. (I know that I have no training in this and I guess that I'm looking at this too literally.)