PDA

View Full Version : GEDMATCH: LIVINGDNA vs 23andme & AncestryDNA



Hurricane
03-26-2018, 08:08 PM
Am i the only one who noticed a slight difference between the results from the LivingDNA kit and the AncestryDNA/23andme kit(s)?
23andMe and Ancestry agree with each other and give similar results.
:\

https://sd-1.archive-host.com/membres/up/1057594130/Others/comparisons2.png

kingjohn
03-26-2018, 08:35 PM
can you post also your eurogenes k13 values from all 3 company ?
kind regards
Adam

p.s
i think the reason is because living dna use a different chip gsa chip that is different from the chip of ancestery dna
and also from {23and me v3,v4 chip }in case this your version .
23and me v5 also use gsa chip like living dna .....

RobinBMc
03-26-2018, 09:08 PM
Well, they use different chips which means there's slightly different raw data, so I would expect there to be slightly different results.

Which 23andMe chip did you test on? These are the total amounts of SNPs for their last few chips: 684,713 (V5); 602,352 (V4); 967,000 (V3). LivingDNA say they have a total of 674,000 SNPs and AncestryDNA include 668,857 SNPs. These differences could account for the small differences in your gedmatch results.

Hurricane
03-27-2018, 04:34 PM
can you post also your eurogenes k13 values from all 3 company ?
kind regards
Adam

p.s
i think the reason is because living dna use a different chip gsa chip that is different from the chip of ancestery dna
and also from {23and me v3,v4 chip }in case this your version .
23and me v5 also use gsa chip like living dna .....

Here it is.


Population
LIVINGDNA
ANCESTRYDNA
23ANDME


North_Atlantic
3.55
2.04
2.12


Baltic
1.25
1.17
1.12


West_Med
4.09
4.61
4.74


West_Asian
-
-
-


East_Med
7.72
9.05
8.86


Red_Sea
12.56
12.63
12.54


South_Asian
1.76
1.48
1.42


East_Asian
0.31
0.17
0.15


Siberian
-
-
-


Amerindian
0.21
0.19
0.16


Oceanian
0.76
0.82
0.93


Northeast_African
19.48
19.64
19.79


Sub-Saharan
48.30
48.19
48.17





Well, they use different chips which means there's slightly different raw data, so I would expect there to be slightly different results.

Which 23andMe chip did you test on? These are the total amounts of SNPs for their last few chips: 684,713 (V5); 602,352 (V4); 967,000 (V3). LivingDNA say they have a total of 674,000 SNPs and AncestryDNA include 668,857 SNPs. These differences could account for the small differences in your gedmatch results.

Im on the V3 chip.
Then which one is the most accurate?

Helgenes50
03-27-2018, 04:40 PM
You have a higher coverage with V3

kingjohn
03-27-2018, 04:51 PM
Here it is.


Population
LIVINGDNA
ANCESTRYDNA
23ANDME


North_Atlantic
3.55
2.04
2.12


Baltic
1.25
1.17
1.12


West_Med
4.09
4.61
4.74


West_Asian
-
-
-


East_Med
7.72
9.05
8.86


Red_Sea
12.56
12.63
12.54


South_Asian
1.76
1.48
1.42


East_Asian
0.31
0.17
0.15


Siberian
-
-
-


Amerindian
0.21
0.19
0.16


Oceanian
0.76
0.82
0.93


Northeast_African
19.48
19.64
19.79


Sub-Saharan
48.30
48.19
48.17






Im on the V3 chip.
Then which one is the most accurate?

from what i see living dna overestimate your north atlantic and east asian
and underestimate your east med
other than that most of the other components are stable in eurogenes k13 with all 3 companies

Angriff
03-27-2018, 05:26 PM
Well, they use different chips which means there's slightly different raw data, so I would expect there to be slightly different results.

Which 23andMe chip did you test on? These are the total amounts of SNPs for their last few chips: 684,713 (V5); 602,352 (V4); 967,000 (V3). LivingDNA say they have a total of 674,000 SNPs and AncestryDNA include 668,857 SNPs. These differences could account for the small differences in your gedmatch results.

The big difference isn't so much the number of SNPs but the overlap in which SNPs they are testing. Ancestry and 23andMe have a fairly high overlap while LivingDNA tests a very different set of SNPs. The overlap is only something like 180k SNPs or something on that scale. Thus these models built with Ancestry and/or 23andMe data sets will be less accurate with LivingDNA data due to the lower sampling of SNPs that will be modeled and accounted for.

RobinBMc
03-28-2018, 03:32 PM
Im on the V3 chip.
Then which one is the most accurate?

That really depends on the individual. I found LivingDNA to be the most accurate for me, but many others don't.

jshook
04-10-2018, 03:36 AM
22578

Mine vary greatly too.

Xtian
04-19-2018, 09:41 AM
Mine vary as well. Im Russia from Krasnodar region.

Although the numbers dont look huge, FTDNA basically aligns me with Ukrainian. The only consistency is that I have relatively high West Asian, double the average for Ukrainians and Russians.

The effect is my FTDNA shows me as a distance as 5 to Ukrainian, with 92% Ukrainian populations whilst my LDNA result is a distance of 9 to Ukrainian with very mixed populations

22668

Mike_G
04-21-2018, 02:34 PM
That really depends on the individual. I found LivingDNA to be the most accurate for me, but many others don't.

I just took a look at your ethnicity results. Living DNA did a great job for you, especially considering your background outside of the UK. Are your 23andme results phased with another relative?

RobinBMc
04-21-2018, 03:47 PM
I just took a look at your ethnicity results. Living DNA did a great job for you, especially considering your background outside of the UK. Are your 23andme results phased with another relative?

No, I'm the only one who took the 23andMe test so far.

fostert
04-21-2018, 04:36 PM
GEDMATCH calculators show the number of SNPs used to derive the ethnicity estimate. For e.g., with my Ancestry v1, 23&me v4 and LivingDNA kits with Eurogenes K36, I find:

Ancestry - 159901 SNPs
23andme - 147243 SNPs
LivingDNA - 52398 SNPs

Above all, its important to remember what these calculators are telling you: they actually do not tell you your ethnicity, but rather only how similar you are in terms of shared SNPs to their reference populations. The big assumption/big leap that people make is that a bunch of SNPs unique to a subset of the reference pop are indicating the "ethnicity" of that subset, and thus your own. Take that for what its worth - an indirect estimate of your ethnicity that rests on a major assumption.

kingjohn
04-21-2018, 08:14 PM
living dna overlapp well with 23and me v5 in terms of number of markers
with all the rest{ ancestery v1, 23and me v3v4 , ftdna , my heritage } it overlaps pretty bad ..

Xtian
04-23-2018, 09:07 AM
GEDMATCH calculators show the number of SNPs used to derive the ethnicity estimate. For e.g., with my Ancestry v1, 23&me v4 and LivingDNA kits with Eurogenes K36, I find:

Ancestry - 159901 SNPs
23andme - 147243 SNPs
LivingDNA - 52398 SNPs

Above all, its important to remember what these calculators are telling you: they actually do not tell you your ethnicity, but rather only how similar you are in terms of shared SNPs to their reference populations. The big assumption/big leap that people make is that a bunch of SNPs unique to a subset of the reference pop are indicating the "ethnicity" of that subset, and thus your own. Take that for what its worth - an indirect estimate of your ethnicity that rests on a major assumption.


Why are LDNAs SNP count so low? And does it matter?

RobinBMc
04-23-2018, 03:34 PM
Why are LDNAs SNP count so low? And does it matter?

I guess it depends on which SNPs included in their test are a part of that calculator's reference panel? LivingDNA say their test includes over 650,000 SNPs, which is about standard. But if their chip covers different SNPs (different sections of the full genome) then the overlap it has with other tests/the reference panels used for admixture could be smaller.

fostert
04-23-2018, 04:22 PM
Why are LDNAs SNP count so low? And does it matter?

LDNA has a customized chip built to test for SNPs that their own scientists have determined give the best representation of a person's ethnic ancestry. Their selection of SNPs will not necessarily be the same as Ancestry's, or 23andme's selection. The reference panels in the GEDMATCH calculators were constructed from people who tested with Ancestry v1 or 23andme's (v3 or earlier?) SNP set, and so it is not surprising that LDNA tests overlap with only 1/3 of them. Even so, LDNA kits will still give you just as valid a result as the other two.

You should take results for each of your Ancestry, 23andme and LDNA and average them for each region, weighted by the numbers of SNPS. The standard deviation of the three measurements then gives you the uncertainty, or error, in each region estimate.
E.g. my K36 results for East Central Europe: Ancestry v1 = 11.09%, 23andme v4 = 11.12% and LDNA=10.85%. Weighted average by the number of SNPs = 10.98%. Standard Deviation = 0.36%. This is a Signal/Noise ratio of 10.98/0.36=31 - a good, strong real component of my ethnicity.

As an alternate example, consider my AmerIndian percents: 0.32%,0.52%,0.0%. The Signal/Noise=1.2. In experimental science anything less than S/N=3-5 is considered noise; therefore I can confidently say that the GEDMATCH calculators cannot measure my AmerIndian component, if I even have one (AFAIK, I don't).

Xtian
04-24-2018, 08:29 AM
Mine vary as well. Im Russia from Krasnodar region.

Although the numbers dont look huge, FTDNA basically aligns me with Ukrainian. The only consistency is that I have relatively high West Asian, double the average for Ukrainians and Russians.

The effect is my FTDNA shows me as a distance as 5 to Ukrainian, with 92% Ukrainian populations whilst my LDNA result is a distance of 9 to Ukrainian with very mixed populations

22668



Ok so you suggest I take an average of my two K13 weighted in favour of FTDNA?

My FTDNA basically aligns me fairly close to Ukrainian whilst LDNA suggest I am a little more mixed. On the actual websites, both LDNA and FTDNA give me around 5% Turkish which I think its linked to my above average West Asian. But LDNA gives me nearly 10% Italian which is really weird.

redeyednewt
05-14-2018, 06:11 PM
I found living DNA to be accurate in some ways and interesting but I did it just to get the YDNA and MT DNA results for my dad, and paternal grandmother (his mother), at a far less expensive price than you would pay for it via Family Tree DNA.