Friday, July 14, 2006
In September of last year, Bruce Lahn's group at the University of Chicago published two articles in Science arguing that two genes which, when mutated, cause microencephaly had recently been (and possibly currently are) under selection in humans. The phenotype under selection was (and still is) unclear, but the fact that mutations in the genes lead to reduced brain size led to a number of hypotheses. Criticisms of the work on moral or political grounds were made, but no one really questioned their actual results. Until now.
The evidence for selection in the two papers based on a couple observations: first, the extent of linkage disequilibrium around an allele is determined primarily by the age of the allele (i.e. when it first mutated from the ancestral state). Second, if an allele is under positive selection, it will probably be at a higher frequency than most alleles "of it's age". So high frequency + long linkage disequilibrium = evidence for selection.
Of course, other factors play a role in the length of LD around an allele and its frequency, the most important of which is arguably demography-- different models of population growth and bottlenecks lead to different patterns of variation in the genome. Lahn's group simulated a number of different demographic scenarios, and conclude that realistic neutral scenarios could not account for the patterns they see.
The current issue of Science has a technical comment on the two papers by Currat et al. which considers some additional demogrphic models. In these models, the patterns in the data are perfectly consistent with neutral evolution. However, as is noted in Lahn's group's response, one could come up with a neutral model to create just about any pattern of sequences. The real question is whether the parameters in that model are plausible.
The plausability of a given parameter in this kind of model is not something I can judge offhand (though perhaps someone else can), but Currat et al. don't cite a single paper to justify their choices, while Lahn's response includes a number of citations to support their decisions and gently suggest that Currat et al. are out of their element. In my book, citations and evidence trump unsupported parameters, so I don't think the case for selection at the two miroencephaly loci is really in question.
Ideally, Lahn's two loci would have turned up in the Voight et al. scan for selection in the human genome, which was fairly robust to different demographic scenarios due to their use of a non-parametric statistic. But their approach had low power for alleles of lower frequency (like those under selection in ASPM and MCPH1), and with a limited sample size of 90 people, they didn't stand out.
So what will resolve this "controversy"? Well, for some people who think the case for selection on lactose tolerance is unfounded, nothing. Otherwise, scans like those by Voight et al. with larger sample sizes should increase the statistical evidence for selection, and functional studies will hopefully elucidate the phentype under selection, making the case complete.
UPDATE: RPM at Evolgen has more.