A Question About Genetically Tracing One's Ancestry

Over at Pandagon, Pam's aunt had her ancestry traced using mitochondrial DNA ('mtDNA'). The results given to her aunt were that she is:

51% Sub-Saharan Africa (she didn't get any information as to country)
37% European (we have known ancestry in Ireland and England)
12% Native American (likely a NY or Connecticut tribe; we thought Shinnecock, but have not been able to confirm)

How is it even possible to conclude that one is 'percent-anything' based on a mtDNA haplotype? Is there something about the test that I'm missing here?

One really can't say that an mtDNA haplotype is 50% X and 50% Y, since mtDNA does not recombine. In other words, it's one big gene that is vertically transmitted from mother to daughter. The only changes are due to mutation.

What this means is that mtDNA can't be 'part-anything.' What one can say is the likelihood that a given mtDNA haplotype is from a certain geographic region or ethnic heritage (note: I said "ethnic", not racial). As far as I can tell, one of two things are happening with the report Pam's aunt received:

1) 51% of those who possess her aunt's haplotype are from Sub-Saharan Africa, 37% from Europe, and 12% are Native American.
2) 51% of the single nucleotide polymorphisms (SNPs; variable bases of DNA--roughly 99% of the total mitochondrial DNA is identical among all humans) are found in Sub-Saharan Africans, 37% in Europeans, and 12% in Native Americans. Actually, it's probably more complex than that. For a given SNP, X% are found in geographic group A, Y% in group B. One then derives a composite likelihood for all the SNPs.

In other words, aren't these percentages a statement of ignorance about human migration and mitochondrial evolution (as well as the consequence of homoplasy), rather than gene flow? Or is there something obvious I've overlooked?

Just for fun, let's assume I'm correct. It is very interesting how it is so difficult to assign a clonal lineage to a given geographic area. There are two reasons for this:

1) Human migration. As an example, imagine if we didn't know that there were large waves of European immigrants to America (bear with me, it's a thought experiment). We would misidentify the geographic background of many samples.
2) Recurrent mutation. mtDNA is screwy. Certain kinds of mutational changes occur much more often than others (and mtDNA was picked because it has a high mutation rate to begin with). Thus, it is possible that the same exact mutation will independently evolve in two separate geographic groups.

Again, am I missing something obvious here?

an aside: If nuclear DNA were being used, and not mtDNA, we could talk about people being genetic composites. This is because of recombination: you are a random shuffling of your parents' genes. Consequently, if you were 1/128 group "X", if we screened enough genes, we might detect that (the same likelihood argument applies to different alleles of gene that applies to the mtDNA SNPs).

More like this

The company she points to does parental ancestry as well as Maternal. My guess is that she did that on instead of mtDNA. But their FAQs also say....

"Can these tests determine "admixture" or percent ancestry? [Back to top]
No, our tests do not determine overall admixture. Our tests are lineage-based tests that trace the ancestry of maternal and paternal lineages only. These lineages may or may not be of recent African origin and should not be confused with percent ancestry. Our tests are different from the admixture tests that provide an estimate of overall continental genetic ancestry (European, Native American, and west African) in that we can determine where shared lineages are in particular regions within Africa."

So it's a little unclear what they are testing.

There is a copy of the mitochondrial genome on chromosome 17. You can read about this here. This would give two additional copies of mitochondrial DNA that could (in theory at least) be amplified and confuse the results.

I've seen people doing these types of calculations with nuclear DNA many times. I agree with the issues you raise, particularly with respect to mtDNA.

However, a few years ago I saw some results where an individual had heterozygous mtDNA. I've forgotten all the details, but I remember that mtDNA from additional family members was available and the hypothesis of the people who did the study was that some paternal mtDNA can be transmitted, but usually it is overwhealmed by the maternal mtDNA - there were a bunch of discussion about mtDNA copy numbers at various stages. But there was no evidence for recombination between the maternal and paternal mtDNA.