Signals of recent positive selection, words vs. figures

Dan MacArthur already posted some of the supplementary figures from Signals of recent positive selection in a worldwide sample of human populations, but he didn't put up one that I thought was really striking. The text:

First, there is extensive sharing of extreme iHS and XP-EHH signals between Europe, the Middle East, and Central Asia, while overlap between other regions is much more limited. In fact, 44% of the genomic segments in the 1% tail of iHS in Europe fall in the 5% tail for both the Middle East and Central Asia (89% are shared between Europe and at least one of these two), while only 12% of European signals are present in East Asia by the same criterion. Second, XP-EHH signals seemto be shared on a larger geographic scale than iHS signals.

Below the fold the figures. Rather stark.


iHS

i-2492b15b91af5c66aebce935eaabf105-fig4ihs.png


XP-EHH

i-b746b2b030115b99b902f7edb01854d9-fig5xpehh.png

More like this

It gives me no pleasure to do this, but sometimes even friends let loose with such jaw-droppingly bad arguments that it is impossible for me not to redirect a bit of the old Respectful Insolence in their direction.
Obama is going to rouse the ire of the religious right yet further: he wisely opted not to endorse Jesus while giving a speech on economics by having a Christian symbol on the lectern c
I don't usually announce these sorts of things on the blog, but since Randy is a long-time friend, colleague and fellow Habs fan, I just had to make an exception.
John Hawks & Daniel MacArthur have already pointed to a new paper in Genome Res

Nice pics - yep, I should have posted 'em. :-)

It's worth spelling out, though, that the stark contrast between African and non-African populations in the second graph is partly an artifact of the XP-EHH algorithm. The algorithm requires the use of an outgroup reference population, and the authors used Bantu as the reference for all non-African pops and Europe as a reference for both African. The authors thus say in the supp methods that "the non-overlap between African and non-African populations by XP-EHH is not meaningful."

That's not to say that there aren't striking differences in selective signatures between Africans and non-Africans, of course (you can still see that in the iHS test, which doesn't use a reference population) - I just wanted to be sure no-one over-interpreted the second graph...

What does this mean?

Thanks for clarifying that point Dan. As we say in the paper, the need for a ref. population in the XP-EHH test means that such a strong divide between Af.-non Af. is indeed an artifact. Though as you say the iHS result backs up the idea that there are few obvious shared sweeps.

By Graham Coop (not verified) on 26 Mar 2009 #permalink

right. i was going to say, "but i assume everyone read the paper!" but then why would i spotlight the sups :-) east vs. west eurasia is still interesting though.

What does this mean?

the description to the figures seems pretty clear, no?