PDA

View Full Version : Why does DNA Land give such strange results?



ianz91
01-30-2019, 09:59 PM
MY DNA Land results 28710


My 23andme results 28712


23andme seems more accurate based on my known ancestry, but why is DNA Land so different? Why such high Balkan? Why is West Eurasian even a thing? lol

digital_noise
01-31-2019, 05:53 PM
DNA Land gives weird results (well, even more weird) with 23 and Me data. I get super high Ashkenazi and Balkan with V5 raw data, bot both of those are absent in the other uploads I have done.

fostert
01-31-2019, 07:21 PM
1) DNA.land imputes (sort of like interpolates, or educated guess) your raw data to fill in the missing SNPs not measured by your vendor's bead chip. MyHeritage, Geneplaza, Gencove, etc all do the same thing. The result is stretches of ethnicity that are not actually read but are guessed to be there.

2) DNA.land has a very different reference population than the other vendors, made up of different people (some probably with fully sequenced genomes) to compare the imputed data to. This different reference population necessarily means different geographic regions and breakdowns for comparable accuracy to the other vendor's regions.

mildlycurly
01-31-2019, 11:54 PM
DNA Land gives weird results (well, even more weird) with 23 and Me data. I get super high Ashkenazi and Balkan with V5 raw data, bot both of those are absent in the other uploads I have done.

My 23andme upload gives me 2.8% Indus Valley (known Romani ancestry) whereas my Ancestry upload gives me none.

As 23andme uses fewer SNPs than Ancestry, it could be picking up on more distant ancestry. Run it through Jtest on GEDmatch and see if you get more than 4% Ashkenazi. That means it's likely real, if distant.

digital_noise
02-01-2019, 04:22 AM
Ive already done that. I might have distant Ashkenazi as I get trace amounts of 1-3% across multiple platforms, but DNA Land gives like 10%, which is absurd

ianz91
02-01-2019, 01:37 PM
Ive already done that. I might have distant Ashkenazi as I get trace amounts of 1-3% across multiple platforms, but DNA Land gives like 10%, which is absurd

I get around 3% Ashkenazi on jtest and have no Jewish ancestry. The creator of the jtest said himself that European descended people to basically ignore it if you score <5% Ashkenazi

digital_noise
02-01-2019, 10:42 PM
I get around 3% Ashkenazi on jtest and have no Jewish ancestry. The creator of the jtest said himself that European descended people to basically ignore it if you score <5% Ashkenazi

yea, I never took JTest seriously. In Fact, I suspect that any Ashkenazi signal I see is related to southern Italian Levant admix.

RobinBMc
02-02-2019, 05:59 PM
1) DNA.land imputes (sort of like interpolates, or educated guess) your raw data to fill in the missing SNPs not measured by your vendor's bead chip. MyHeritage, Geneplaza, Gencove, etc all do the same thing. The result is stretches of ethnicity that are not actually read but are guessed to be there.

2) DNA.land has a very different reference population than the other vendors, made up of different people (some probably with fully sequenced genomes) to compare the imputed data to. This different reference population necessarily means different geographic regions and breakdowns for comparable accuracy to the other vendor's regions.

23andMe V5 chip uses imputation as well.

Dorkymon
02-02-2019, 06:26 PM
Works pretty well for me with 23andme v4.
Compare to what I've got in the signature.

West Eurasian 98%
Balkan 53%
North Slavic 38%
Northwest European 3.9%
Ambiguous 3.9%

East Asian 1.7%
Ambiguous 1.7%

Telfermagne
02-26-2019, 01:44 AM
I've not had any issues with DNA.Land. They've been pretty consistent for me regardless the raw data used (23andMe v2 to V3, 23andMe v5, FTDNA, MyHeritage, and AncestryDNA; LivingDNA isn't supported regrettably).29061