tag:blogger.com,1999:blog-8480427453110572235.post5507984793334883583..comments2008-06-27T12:48:59.234+10:00Danielhttp://www.blogger.com/profile/07276690118219000204noreply@blogger.comBlogger3125tag:blogger.com,1999:blog-8480427453110572235.post-98434779627140122008-06-27T12:48:00.000+10:002008-06-27T12:48:00.000+10:00I just took a quick look at the paper. I'm not sure that the regression model is very convincing, it probably helps deal with some of the correlation but I don't know how much. Also the effect use of the control alleles depends on whether those alleles are well matched in terms of frequency etc. So I'm not overly reassured, it is really hard to convince yourself about these kind of results.<BR/><BR/>As you know, even if this allele has indeed been under selection there are probably many other environmental variables that correlate with 'migration distance' so it is hard to assign causation even if the correlation is real.Ghttp://www.blogger.com/profile/05140268044719927398noreply@blogger.comtag:blogger.com,1999:blog-8480427453110572235.post-61071574767107134862008-06-26T14:31:00.000+10:002008-06-26T14:31:00.000+10:00Hi G,<BR/><BR/>This is a great point. (For other readers: nearby populations may have similar allele frequencies simply because they share more recent common ancestors, a problem known as auto-correlation.) The authors did try to account for this possibility in two ways. <BR/><BR/>Firstly, they attempted to correct for auto-correlation using a network regression model; the association shown in that graph is still significant (P < 0.001) after this correction. I don't know how convincing that correction method is - any thoughts?<BR/><BR/>Secondly, they checked frequency distributions for 128 alleles summarised in Cavalli-Sforza's History and Geography of Human Genes, and found none with a consistent association with migration distance. Of course, the data-sets available now would allow a much more rigorous assessment of this question: it would be very interesting to see if SNPs tagging this variant (assuming there are any) sit in the empirical tail of the distribution of genome-wide SNPs for correlation with migration distance from Africa.Danielhttp://www.blogger.com/profile/07276690118219000204noreply@blogger.comtag:blogger.com,1999:blog-8480427453110572235.post-9966611393850719822008-06-26T13:08:00.000+10:002008-06-26T13:08:00.000+10:00I've not read the paper the graph is taken from but I suspect that many neutral alleles could show these kind of patterns. There are likely to be relatively few independent points on this graph due to the shared history of most human populations. Do they look for such patterns at any control SNPs? If not, this graph is not very convincing.Ghttp://www.blogger.com/profile/05140268044719927398noreply@blogger.com