Wednesday, March 28, 2007
PLoS One has a new article on detecting recent natural selection in the human genome. As opposed to previous studies mentioned on this site , this study looks for fixed differences between populations, rather than those on their way to fixation. This is a potentially very important set of differences between populations, and one that is as of yet unexplored.
My inital reaction to this study was lukewarm, but their approach isn't biased in any way, rather it's just not particularly rigorous-- the statistic they use is based on blocks of homozygosity, an approach which is strongly influenced by demographics. They note themsevles that, among the candidates for being under selection, "a relatively large proportion...are false positives". This makes it particularly difficult to judge any single gene as being under selection from this dataset. However, some of the genes they report as being under selection are worth mentioning.
First, when comparing African and European-origin samples (the data used are from the HapMap), they find a number of pigementation related genes, as well as genes involved in skeletal development. These genes could be involved in the notable phenotypic differences between Western European and Nigerian populations. Some gene involved in fertility also show up, as do immune-reponse related genes. But as mentioned before, it's tough to make any strong statements with an approach like this.
How would I attack this problem? The limited amount of data, along with some amount of prior information, cries out for a Bayesian approach. Plus, the fact that the chimpanzee genome is also available could be used to one's advantage. This paper is a decent first strike at the problem of identifying important population-specific adaptations, but expect much more rigour in the future...
UPDATE: I've started a discussion on this paper on the PLoS One website (see the first link in the post). Feel free to join in.