ABBA and BAAB test of admixture (D test)

Question

[vicky]

Hi All

If someone has used this test of admixture of population, can help me in solving one doubt.

I want to know that to do this ABBA or BABA test, does sequences needs to be mapped (bam file) against the outgroup species only. http://www.popgen.dk/angsd/index.php/Abbababa

Best

Answer 1

[lh3]

Would be theoretically better to do that, but usually not critical when you have high depth and a decent reference genome. Definitely the data should be mapped to the same reference genome.

[vicky]

Thanks for the reply. Just to be sure say I have African humans, Asian humans, Neanderthal and Chimpanzee. So in this scenario, I will map the reads of Asian, African, Neanderthal to Chimp genome. Automatically the Chimp becomes the ancestral state. But I cannot map my reads of Asian,African and Neandethal to human genome and use the test? That is not possible?

[Gabriel R.]

But why don't you just map to the human reference and use the EPO data to infer the ancestral allele?

BTW, D-stats are not robust measures of admixture. They just tell you that there is an imbalance between both sides of the tree but do not tell you whether this is due to genome quality, modern human contamination or ancient substructure.

[vicky]

Thanks for reply. I just used the example here for humans I have different species. I just had the doubt that the outgroup species in the software mentioned above cannot be used inside the three species say (A,B:C) which says it should be like this (A,B:C,O).

Now the outgroup will be ancestral state, C will be derived and you will check if geneflow is between A,B and C so the A,B,C should be mapped to only outgroup which will act as a reference genome only?Or it is not required.

That was my question.

[lh3]

You can map reads to the human genome.

Answer 2

[Gabriel R.]

No unless you suspect major reference bias. So if you can get decent derived and ancestral allele calls, I would not worry about it.