PDA

View Full Version : Question about 23andme Discrepancy



cyna
11-22-2013, 04:29 PM
Ok, so I have tested my 23andme raw data pretty widely, and except for 23andme, I get the same result at all places:

Eurogenes K15 = 52.74% SSA, 9.18% East Asian + Native American, 38.08% Euro

Eurogenes K13 = 52.4% SSA, 9.54% East Asian + Native American (East Asian), 38.06% Euro

Eurogenes K12 = 52.18% SSA, 9.07% East Asian + Native American, 38.75% Euro

Eurogenes K12b = 52.28% SSA, East Asian 9.24%, 37.96% Euro

Eurogenes K11 = 52.2% SSA, East Asian + Native American, 38.5% Euro

Eurogenes K10 = 52.2 SSA, 9.3% East Asian + Native American, 38.5% Euro

Eurogenes K9 = 52.2% SSA, 9.32% East Asian + Native American, 38.48% Euro

Eurogenes K9b = 51.27% SSA, 10.34% East Asian, 38.39% Euro

McDonald = 53.1% SSA, 8.8% East Asian, 38.1% Euro

FTDNA = 52.8% SSA, 9.1% East Asian, 38.1% Euro

Dodecad World9 = 51.55% SSA, 9.78% East Asian, 38.67% Euro

Dodecad K7 = 52.5% SSA, 9.03% East Asian, 38.47% Euro

Dodecad K12 = 53.04% SSA, 9.12% East Asian, 37.84% Euro

Harrapa = 52.83% SSA, 8.84% East Asian, 38.33% Euro

DNA Tribes = 51.7% SSA, 9.2% East Asian, 39.1% Euro

These results were all generated with 23andme raw data. As you can see, they are all, more or less, within a tight 1% range.

----------------------------------------------------------------------------------------

23andme = 58% SSA, 8% East Asian, 34% Euro

Honestly, I am quite surprised by this result since it deviates from all of the other results by a statistically meaningful amount. What do you guys think? Also, with the update coming at 23andme, do you think it like that my 23andme results will converge with the pack?

Dr_McNinja
11-22-2013, 05:04 PM
I haven't seen such consistent results before. Count yourself lucky I suppose. Some of ours vary wildly with each service. There's a significant difference in my results even between Eurogenes calculators.

Ann Turner
11-22-2013, 05:24 PM
One difference might be the populations used as reference samples. 23andMe has added quite a few samples based on the Where Are You From? survey.

cyna
11-22-2013, 05:27 PM
They also just added a whole host of new populations and are undertaking a re-computation. Chance that 23andme will converge on the rest? The population sets are the same ones used by the others from what I understand.

Diverclic
11-22-2013, 05:31 PM
I haven't seen such consistent results before. Count yourself lucky I suppose. Some of ours vary wildly with each service. There's a significant difference in my results even between Eurogenes calculators.

I agree (and I confirm) : your results are amazingly similar with the different calculators.

cyna
11-22-2013, 05:34 PM
Except for 23andme :\

S9 H9
12-02-2013, 12:39 AM
As has been pointed out before, the 23andMe AC is not intended to do exactly the same thing as the Eurogenes or Dodecad utilities.

AC is not just counting SNPs. It looks at windows of SNPs in an attempt to assign some identity to them, an identity within a timeframe that has historical meaning for people today. Then runs of windows are smoothed, which will tend to reduce traces of older ancestry. The design goal of the algorithm is summarized in the "500 years" phrase which often appears to be misunderstood.

Even with all of that, your 23andMe results do not differ much from the admixture calculators.

Sein
12-02-2013, 02:29 AM
I second S9 H9 on this. Also, 23andMe has some East African samples to power the "Sub-Saharan African" category. This probably explains the higher African percentage at 23andMe, since some of your West Eurasian ancestry is being absorbed into that category.