Some questions about my South Asian Ancestry

[misinformedgeneticist]

So I took a 23andme test a while back and got 92.6% South Asian, 5.3% West Asian, and 1% European.

But when I recently analyzed my results using PuntDNAL k13, I got this:

West_Asia 39.08
NE_Europe 13.23
Americas 2.21
Siberia 1.06
Oceania 0.17
South_Asia 33.98
NE_Asia 1.65
East_Africa -
SE_Asia 0.20
SW_Europe 6.07
SW_Asia 0.98
West_Africa 0.91
South_Africa 0.46

I was playing around with different calculators and projects and I'm getting similar results. What's up with the major discrepancy? Why am I getting a much lower South Asian percentage using Gedmatch? Is Gedmatch less accurate? Or does it show higher East Asian and European results because Indians just have a recent common ancestor with the two populations? I understand genealogy is a little bit more complicated than what most people think, but can someone shed light on the discrepancy with the results?

As a side note, I don't know anyone in my family from West Asia. I assume Pakistan is classified as South Asia and I know that my grandparents lived in Pakistan prior to India and Pakistan breaking up.

To make matters even more confusing, when I played around with the confidence interval in 23andme, I got 72.7% south asian, and 0.2% European with no East Asian whatsoever. So, why am I getting no East Asian with a 90% confidence interval, yet I am somehow getting European, using 23andme?

Thanks!

[msmarjoribanks]

Gedmatch and 23andMe (and the other big commercial services) do different things.

23andMe tries to identify recent ancestry, so if your family is all from India, it generally should give you 100% South Asian, and ideally break it down into regions. I'm not knowledgeable about how good they claim to be at regions within South Asia, though. The West Asian is probably the same reason I get 2% Caucasus and 2% Iberian on Ancestry -- sometimes it recognizes components that are within your other results as if they were from a separate ancestry. It's normal for there to be DNA patterns (components) in South Asian ancestry that look West Asian (largely due to more distant historical ancestry). This could also be because their South Asian samples blend together people from different regions to some degree and your particular ancestry has a bit more of a West Asian look (but 5% isn't all that high).

The different Gedmatch calculators start with certain types of components that they attempt to recognize and then identify probable ancestries or mixes based on the combination of those components (the oracles). So everyone gets a big mix and they tell you the mix, not just the conclusions based on that. Instead, they show a bunch of different populations that fit.

I think 23andMe is more trustworthy than a Gedmatch oracle (Gedmatch oracles show numerous possibilities and you shouldn't take any particularly seriously), but seeing the Gedmatch components is really interesting and can kind of help understand why the commercial tests conclude as they do.

One thing I'd recommend for better understanding how Gedmatch works is to look at the spreadsheet that goes with the calculator if they have one and looking at the averages for backgrounds similar to yours.

I don't know what your specific background is, but for just one example, the average Punjabi in their sample pool for the PuntDNAL K13 gets:

41.5% West Asia
13% NE Europe
38% South Asia

and then smaller amounts of many other components

The PuntDNAL global K13 is supposed to work for people of any ancestry, so has very very broad categories, so I expect percentage of the "West Asia" component (which doesn't assume actual West Asian ancestry) helps identify different parts of South Asian from each other.

Kind of similarly, there's a huge NE Europe category and a huge SW Europe, but no NW Europe -- so people of, say, primarily English background get pegged there based on their combination of the other things, even if they don't really have much that we'd normally call NE or SW Europe -- it's old ancestry. (That's me -- my top Oracle result is English, which is reasonably accurate, and my main components are 48% NE Europe, 34% SE Europe, and 10% West Asia -- I also have no real West Asian ancestry.)

Hope that helps some. The Gedmatch calculators can be really interesting once you play around with them for a while and get familiar with how they work.

[msmarjoribanks]

Here's a discussion of some ancient migration patters to South Asia that show both West Asian input and steppe input (which may be what the NE Europe is reflecting): https://www.gnxp.com/WordPress/2018/...tic-landscape/

It's probably more detailed than what you currently need, but might be interesting. Razib Khan has a podcast where I know he's complained a bit about the way commercial services lump South Asian together when it's quite varied, and he has a project that's supposed to examine it more closely:

http://www.razib.com/wordpress/?cat=2779