Dr. Rob weighs in on the lack of a relationship between mitochondrial DNA (mtDNA) polymorphism and population size. To recap: DNA polymorphism should predict population size -- larger populations will have more polymorphism than small populations. Additionally, comparing different measures of polymorphism allow us to infer ancient demographic events. But natural selection can also shape DNA polymorphism. Positive (or Darwinian) selection will drive an advantageous allele to fixation, taking linked neutral variants along with it. This selective sweep will decrease polymorphism in a region surrounding the site under selection; the size of the region affected depends on the strength of selection and the recombination rate in the region.
And that's where the mtDNA story begins. Because there is little/no recombination in the mitochondrial genome, a selective event at any site will sweep out variation across the entire molecule. Any pattern of mtDNA polymorphism left by population size will be swamped out by these selection events. Dr. Rob refers to this as 'genetic draft'. It's probably not a good idea to use mtDNA alone to study historical population sizes, just like it's a bad idea to use any single marker to infer demographic events.
But I have covered all of this before. What's different this time is something Dr. Rob points out: the ability of selection to affect DNA sequence polymorphism also depends on population size. Adam Eyre-Walker brings this up in his review of the mtDNA polymorphism paper. In humans, autosomal, mtDNA, and Y chromosome polymorphisms all pretty much suggest the same ancestral population size. Does that mean that mtDNA (and Y chromosome) polymorphism is ok for inferring ancestral demographic events?
Not so fast, my friend. You see, the efficacy of natural selection depends on population size. Larger populations will be able to fix more mildly beneficial mutations and purge more weakly deleterious mutations than small populations. In small populations, more mutations are effectively neutral, so genetic drift will play a larger role in fixing mutations in small populations compared to larger populations. Therefore, if most new mutations are only mildly advantageous or deleterious, polymorphism should reflect population size in small populations, but may not in a large population.
The previous scenario applies to a single locus, and remember that the mitochondrial genome essentially behaves as a single locus. If the selective events on mtDNA are of the nearly neutral variety (ie, they don't confer an immense selective advantage), there will be fewer selective sweeps in small populations. So, for small populations, mtDNA polymorphism may be able to identify ancestral demographic events. But patterns of mtDNA polymorphism shouldn't be compared between smaller and larger populations.
- Log in to post comments
The paragraph right after the italicized one should also be italicized --both comprise a direct quote. Sorry for formatting problem.
I'm not sure I follow you here. You say:
But I have covered all of this before. What's different this time is something Dr. Rob points out: the ability of selection to affect DNA sequence polymorphism also depends on population size. Adam Eyre-Walker brings this up in his review of the mtDNA polymorphism paper. In humans, autosomal, mtDNA, and Y chromosome polymorphisms all pretty much suggest the same ancestral population size. Does that mean that mtDNA (and Y chromosome) polymorphism is ok for inferring ancestral demographic events?
Not so fast, my friend. You see, the efficacy of natural selection depends on population size. Larger populations will be able to fix more mildly beneficial mutations and purge more weakly deleterious mutations than small populations. In small populations, more mutations are effectively neutral, so genetic drift will play a larger role in fixing mutations in small populations compared to larger populations. Therefore, if most new mutations are only mildly advantageous or deleterious, polymorphism should reflect population size in small populations, but may not in a large population.
It is, of course, a basic assumption of population genetics that population size and selection/drift are directly connected. But the Bazin et al. paper that Eyre-Walker reviews questions that connection for the mtDNA data they survey. Gillespie, too, questions the assumption about population size dependency. So, it seems question-begging to respond to Bazin et al. (and Eyre-Walker) that natural selection just depends on population size. Or have I misunderstood your post?
At a neutral locus, polymorphism should be a good predictor of population size. Because we don't know which loci are truly neutral, we need to assay many loci across the genome to infer demographic history (demography should produce a genome-wide effect on polymorphism, whereas selection will be limited to around the locus in question). That should justify not inferring population size from mtDNA sequence alone. What makes mtDNA even worse for such studies is that selection at any site on the molecule will affect polymorphism at every other mtDNA locus.
My point about population size and selection is that the efficacy of the selection events are dependent on population size. So, it's not just as simple as saying mtDNA polymorphism doesn't capture population size. In small populations, selection may not be strong enough to influence mtDNA polymorphism. That means that in small populations (ie, humans) mtDNA polymorphism may be a good tool for estimating population size. But in large populations, selective sweeps will be skew the polymorphism so that it now longer is a good estimate of population size.
But this is all for a single locus. I still maintain that demographic history can be studied if we examine enough independent loci regardless of the population size.