What do you think gives more accurate results... GEDMatch calculators or 23andme?

[Sikeliot]

Some people say that 23andme predicts ancestry from the last 500 years, and that it is therefore more reliable than GEDmatch. Others say GEDmatch suffers less from "calculator effect" and thus gives clearer pictures.

I don't really agree because I've noticed very peculiar things happen with the 23andme calculator. I've seen families where one parent is Sicilian and the other is Irish. The Sicilian parent scores 15% Middle Eastern, while the child scores 0%. I've yet to see a half North European, half SE European (Cretan, Sicilian, Greek islander) score significant Middle Eastern even though full members of said groups score a significant amount.

Likewise, I suspect 23andme underestimates admixture when it is small. I have seen that in people with under 10% of African, the African is greatly underestimated on 23andme.

What do others think? I think due to its algorithms, 23andme ends up giving misleading results, but it is good for telling if a given ancestral component exists... just not necessarily the amounts.

[Gooding]

Some people say that 23andme predicts ancestry from the last 500 years, and that it is therefore more reliable than GEDmatch. Others say GEDmatch suffers less from "calculator effect" and thus gives clearer pictures.

I don't really agree because I've noticed very peculiar things happen with the 23andme calculator. I've seen families where one parent is Sicilian and the other is Irish. The Sicilian parent scores 15% Middle Eastern, while the child scores 0%. I've yet to see a half North European, half SE European (Cretan, Sicilian, Greek islander) score significant Middle Eastern even though full members of said groups score a significant amount.

Likewise, I suspect 23andme underestimates admixture when it is small. I have seen that in people with under 10% of African, the African is greatly underestimated on 23andme.

What do others think? I think due to its algorithms, 23andme ends up giving misleading results, but it is good for telling if a given ancestral component exists... just not necessarily the amounts.

I think they're both useful in their own way.. like 23andMe could give you a reasonable, if broad, estimate of where in the world your ancestors came from before embarking for North America, Oceania or South Africa. Gedmatch is good I think for giving broad results for slightly deeper ancestry.. like the particular ethnic elements of your Italian or Cape Verdean ancestors, how near you plot to Levantines, Hellenes, etc..

[Sikeliot]

I think they're both useful in their own way.. like 23andMe could give you a reasonable, if broad, estimate of where in the world your ancestors came from before embarking for North America, Oceania or South Africa. Gedmatch is good I think for giving broad results for slightly deeper ancestry.. like the particular ethnic elements of your Italian or Cape Verdean ancestors, how near you plot to Levantines, Hellenes, etc..


I think 23andme is better for determining if an admixture exists that is recent, and GEDmatch is better for quantifying that admixture.

[Gooding]

I think 23andme is better for determining if an admixture exists that is recent, and GEDmatch is better for quantifying that admixture.

That right there. Yes. Agreed.

[Sikeliot]

That right there. Yes. Agreed.

I think when you use reference samples that are modern populations and not broad categories like "West Asian" or "West African" or "North European", you lose some of the components in clusters and they get hidden. This happens on 23andme. By having an "Italian" cluster, Middle Eastern gets hidden in there because of the Neolithic affinity of the Italian cluster. Someone who has Iberian and African ancestry will lose some of their African score, as Iberians have an African affinity and it'll suck up some of the admixture. And so on.

[Iloko]

Yes I agree about 23andme underestimating ancestry.

On 23andme I'm only 1.9% European, but on Gedmatch I get 5%+ Euro especially on the Eurogenes calculators.

I even think Ancestry.com gives better and higher estimates for me, because I get 6%(<1%-11% range) Asia Central which I believe is tied in with the European I get on other tests and that seems to be a better and higher estimate than 23andme.

Also I only get 4% Chinese on the 23andme composition, but my mom instead gets 20%. Their algorithm and ethnicity predictors have weird issues.

[Gooding]

Yes I agree about 23andme underestimating ancestry.

On 23andme I'm only 1.9% European, but on Gedmatch I get 5%+ Euro especially on the Eurogenes calculators.

I even think Ancestry.com gives better and higher estimates for me, because I get 6%(<1%-11% range) Asia Central which I believe is tied in with the European I get on other tests and that seems to be a better and higher estimate than 23andme.

Also I only get 4% Chinese on the 23andme composition, but my mom instead gets 20%. Their algorithm and ethnicity predictors have weird issues.

That's for sure. I'm of the opinion that 23andme seriously undercounted my British and Irish scores, while perhaps shrouding some of that under the 22.2% " Broadly Northern European." My 56.5% might be more accurately read as 78.2%.

[Damião de Góis]

The runs on Gedmatch (Dodecad and Eurogenes) are a lot more accurate than 23andme's Ancestry Composition. That goes without saying.

[Green]

I think when you use reference samples that are modern populations and not broad categories like "West Asian" or "West African" or "North European", you lose some of the components in clusters and they get hidden. This happens on 23andme. By having an "Italian" cluster, Middle Eastern gets hidden in there because of the Neolithic affinity of the Italian cluster. Someone who has Iberian and African ancestry will lose some of their African score, as Iberians have an African affinity and it'll suck up some of the admixture. And so on.

This has happened to me, according to 23andMe I am in conservative and 96.7% european

https://lh3.googleusercontent.com/fNQASx8JTUv8t6OX5B0GRsrmsoRpJfggnikUCwi-T3E=w1698-h955-no

https://lh3.googleusercontent.com/r7yyFMsqNdmFJoroeBtvqqTdYatO62edo154s-XnCxI=w1698-h955-no

and at gedmatch I find this :blink:.

https://lh3.googleusercontent.com/bvRpJWhKUJImj2XADT620ygMaPbe7QF5sMW6lCi_gTM=w1698-h955-no

It was a very big surprise.

[Neon Knight]

What people forget, in their enthusiasm, is that the specualtive estimate on 23andMe is only clamied to be at 50% confidence level and even the standard estimate is only 75% confidence. So do the GEDmatch Oracles tell us much more than the 23andMe conservative estimate (90% confidence)?

For me, I get a conservative estimate of 37% British/Irish and 46% Northern Euro. Over a few GEDmatch Oracles I get mostly British Isles with a suggestion of Dutch. So in my case, GEDmatch is considerably more specific, but it doesn't work for everyone. My mum did the AncestryDNA test which was spot-on but GEDmatch said she was German. The acid test is this: if you knew nothing of your ancestry, what would you conclude if you were trusting these tests?

[safinator]

23andMe is a professional company, Gedmatch one the other hand relies on amateurs work which can indeed be interesting and fun but at the end of the day it remains still...amateur.

[Catkin]

The population estimates can jump about a bit on 23andme too. I started off as 60-something percent British and Irish on speculative. Then when my mum got tested and scored in the mid-80 percents I jumped up to about 80%. Shortly after my dad got tested and scored somewhere in the 50 percents, and it obviously confused things because I dropped back to the 60s again. It seems speculative really is quite speculative.

[Green]

23andMe is a professional company, Gedmatch one the other hand relies on amateurs work which can indeed be interesting and fun but at the end of the day it remains still...amateur.

So you think that 23andMe is more accurate?

What if I do a test with another company for example Ancestry DNA be equal, more accurate or less accurate?

[Iloko]

I like Ancestry.coms test a lot. I wouldn't mind blowing off another 100 bucks just to test my dad too.

[safinator]

So you think that 23andMe is more accurate?

What if I do a test with another company for example Ancestry DNA be equal, more accurate or less accurate?

Well a bit more accurate but not perfectly so, frankly i don't think the companies you listed would provide a better breakdown than 23andMe either.

[Green]

Well a bit more accurate but not perfectly so, frankly i don't think the companies you listed would provide a better breakdown than 23andMe either.

It does not make sense in 23andMe say they have nothing of North African or Arabic and the gedmach tell me I have 2.51% Arab and 3.08% North African.

One of the two failed much.

[Gaston]

Relying on only one test is inaccurate.

So you have to use your 23andme results/Ftdna/etc (and understand how/why these results), and other tests available at gedmatch and elsewhere (interpretome, the now unavailable Dr McDonald, etc).
You also have to combine different methods, like ADMIXTURE, PCA plots, IBD sharing etc.


All this, together with being up-to-date in population genetics (many people are still unaware of Lazaridis et al.), will make the best interpretation of your results.


This has happened to me, according to 23andMe I am in conservative and 96.7% european

It was a very big surprise.

23andme relies on how much you match the reference population. Other people from your region might get lower "European"/more diverse results not because they are less native than you but because their genomes happen to not match the reference population.

Some Iberians get 100% European/Iberian by the way. Completely useless information.


It does not make sense in 23andMe say they have nothing of North African or Arabic and the gedmach tell me I have 2.51% Arab and 3.08% North African.


The gedmatch one is detecting signals missed out by 23andme because of the better reference populations.
Your 5%+ MENA is common in Iberia, you have to compare yourself to other Iberians.

[Gooding]

What people forget, in their enthusiasm, is that the specualtive estimate on 23andMe is only clamied to be at 50% confidence level and even the standard estimate is only 75% confidence. So do the GEDmatch Oracles tell us much more than the 23andMe conservative estimate (90% confidence)?

For me, I get a conservative estimate of 37% British/Irish and 46% Northern Euro. Over a few GEDmatch Oracles I get mostly British Isles with a suggestion of Dutch. So in my case, GEDmatch is considerably more specific, but it doesn't work for everyone. My mum did the AncestryDNA test which was spot-on but GEDmatch said she was German. The acid test is this: if you knew nothing of your ancestry, what would you conclude if you were trusting these tests?
I would make a lot of erroneous conclusions.

[Green]

The gedmatch one is detecting signals missed out by 23andme because of the better reference populations.
Your 5%+ MENA is common in Iberia, you have to compare yourself to other Iberians.[/QUOTE]

Do I have Arab or North African features?

https://lh3.googleusercontent.com/xLX6CZZR0639DwN__KEt5N6B2Y0BMxSm3CEjNMn1Pw0=s955-no

[Gaston]

Do I have Arab or North African features?


ADMIXTURE results don't correlate with superficial external traits. You might have disease risks and HLA alleles and haplotypes linked to your higher MENA affinity (that would be absent in a Czech or a Finn).

You should ask yourself if the 40% ANE (Ancient North Eurasians) Native Americans share physical traits with 30%+ ANE Northwest South Asians, 20-30% ANE Caucasians, 10-20% ANE Northern Europeans, and 6-10% ANE Iberians.
You have ANE ancestors, do you have Karitiana traits? No.

[Petalpusher]

It does not make sense in 23andMe say they have nothing of North African or Arabic and the gedmach tell me I have 2.51% Arab and 3.08% North African.

One of the two failed much.

It's one test, you may not have anything over noise levels anywhere else, don't stop at one autosomal. On k36 you score higher than me on French...that's the kind of oddity you can get sometimes, it mixes ancient population, broader groups, with actual countries. Try k23b, most of us have a reading of North African and unexpected things, you ll see how it compares.

ps: and for god's sake copy/paste your results :D

[Green]

ADMIXTURE results don't correlate with superficial external traits. You might have disease risks and HLA alleles and haplotypes linked to your higher MENA affinity (that would be absent in a Czech or a Finn).

You should ask yourself if the 40% ANE (Ancient North Eurasians) Native Americans share physical traits with 30%+ ANE Northwest South Asians, 20-30% ANE Caucasians, 10-20% ANE Northern Europeans, and 6-10% ANE Iberians.
You have ANE ancestors, do you have Karitiana traits? No.

Curious gedmatch these data, a few years back ?, 500 years, 1000 years or 2000 years. So it may be a distant cousin of Saladin.

Clearly, the reconquest of Spain and its repopulation with Christian was not very effective.