Identical triplets 23andMe

[Peterski]

I think there is a chance that their GEDmatch results* would actually be 100% identical (as long as all 3 raw datas from the same company):


https://www.youtube.com/watch?v=rDeyeUAtdrQ

*Or their results in any other admixture calculator that is not on GEDmatch. Anyone has GEDmatch kits of some identical twins or triplets?

[Dick]

makes sense. the police can tell twins apart via dna test too

[Peterski]

makes sense. the police can tell twins apart via dna test too

These are just single mutations, I doubt it could influence ethnicity estimate so much (+/- 10%):

https://geneticliteracyproject.org/2019/03/05/crime-scene-investigators-couldnt-tell-identical-twins-dna-apart-until-now/

[Ljubic]

These are just single mutations, I doubt it could influence ethnicity estimate so much (+/- 10%):

https://geneticliteracyproject.org/2019/03/05/crime-scene-investigators-couldnt-tell-identical-twins-dna-apart-until-now/

Not even that much, my identical sibling is like 98% similiar with me. I could use his Raw data and claim it to be mine.

[Pine]

A proper journalist would've reached out to 23andMe for comment. I'd be interested to hear their explanation.

[talombo]

Someone at the 23andme forums got tested twice with the same v5 chip and got slightly different results,
this is what the Forums Moderator had to say about it:



The differences you are seeing in your Ancestry Composition results are most likely the result of the way in which the feature divides and then assigns sections of DNA. The analysis that Ancestry Composition provides involves segmenting chromosomes into short pieces or windows. We compare the DNA you have in each window to the DNA in the same window in our reference dataset, and assign the DNA in each window to the population whose DNA it's most similar to. Then we process those assignments further, "smoothing" them out. For instance, if you have a long run of assignments from population A, interrupted by an assignment to population B, this process can correct that B to an A. Between twins or in your case, two samples from the same person, not every window will start and end at exactly the same place, which may result in small variations between the results.

[Peterski]

Not even that much, my identical sibling is like 98% similiar with me. I could use his Raw data and claim it to be mine.

Can you post some GEDmatch results (admixture percentages) for both of you, for example from Eurogenes calculators?

[Ljubic]

Can you post some GEDmatch results (admixture percentages) for both of you, for example from Eurogenes calculators?

I will use Eurogenes K13 for the example.
Me (Ancestry):

Population
North_Atlantic 25.47 Pct
Baltic 31.24 Pct
West_Med 16.53 Pct
West_Asian 6.96 Pct
East_Med 16.81 Pct
Red_Sea 1.83 Pct
South_Asian -
East_Asian -
Siberian 0.67 Pct
Amerindian 0.23 Pct
Oceanian -
Northeast_African -
Sub-Saharan 0.27 Pct

Sibling (MyHeritage):

Population
North_Atlantic 25.20 Pct
Baltic 31.62 Pct
West_Med 16.32 Pct
West_Asian 7.19 Pct
East_Med 16.62 Pct
Red_Sea 1.86 Pct
South_Asian -
East_Asian -
Siberian 0.47 Pct
Amerindian 0.47 Pct
Oceanian -
Northeast_African -
Sub-Saharan 0.25 Pct


Also One to One:

Largest segment = 151.8 cM

Total Half-Match segments (HIR) = 3557.2 cM (99.190 Pct)
Estimated number of generations to MRCA = 1.0

47 shared segments found for this comparison.

421469 SNPs used for this comparison.

99.993 Pct SNPs are full identical

Comparison took 0.229 seconds.
CPU time used: 0.047 cpu seconds

We didnt even test with the same company but still very similiar.

[Zoro]

Someone at the 23andme forums got tested twice with the same v5 chip and got slightly different results,
this is what the Forums Moderator had to say about it:

The differences you are seeing in your Ancestry Composition results are most likely the result of the way in which the feature divides and then assigns sections of DNA. The analysis that Ancestry Composition provides involves segmenting chromosomes into short pieces or windows. We compare the DNA you have in each window to the DNA in the same window in our reference dataset, and assign the DNA in each window to the population whose DNA it's most similar to. Then we process those assignments further, "smoothing" them out. For instance, if you have a long run of assignments from population A, interrupted by an assignment to population B, this process can correct that B to an A. Between twins or in your case, two samples from the same person, not every window will start and end at exactly the same place, which may result in small variations between the results.




How STUPID :picard2: so if you are European and your bigger segment A is European and your smaller segment B is W. Asian they will ALTER segment B and make it European. No wonder most people turn out close to 100% European or 100% W. Asian. So basically if your W. Asian segments are any smaller than your European segments they will just erase them. Isn’t that cheating?

[Pine]

How STUPID :picard2: so if you are European and your bigger segment A is European and your smaller segment B is W. Asian they will ALTER segment B and make it European. No wonder most people turn out close to 100% European or 100% W. Asian. So basically if your W. Asian segments are any smaller than your European segments they will just erase them. Isn’t that cheating?

His explanation isn't precise enough. However, 23andMe's smoothing algo lowers their false positive rate, but elevates their false negative rate.

[Zoro]

His explanation isn't precise enough. However, 23andMe's smoothing algo lowers their false positive rate, but elevates their false negative rate.

Well ok It’s one thing as far as lowering false positive to cancel 1 SNP here or there but wiping out whole matching segments of SNPs just because they happen to be smaller than the dominant segment in the window is nuts and of course will increase their false negative rates but I guess they figure a British tester will not complain if they tell him he is 100% European and has no west Asian ancestry :rolleyes:

So I guess unless the British guy has huge W. Asian segments from a great grandfather or grandfather his W. Asian segments stand no chance of showing up

[Pine]

Well ok It’s one thing as far as lowering false positive to cancel 1 SNP here or there but wiping out whole matching segments of SNPs just because they happen to be smaller than the dominant segment in the window is nuts and of course will increase their false negative rates but I guess they figure a British tester will not complain if they tell him he is 100% European and has no west Asian ancestry :rolleyes:

So I guess unless the British guy has huge W. Asian segments from a great grandfather or grandfather his W. Asian segments stand no chance of showing up

Why do you all default to an agenda when it comes to this shit? It's simply because 23andMe prefers a low false positive rate. That's why they have such high general assignments: "unassigned" , "broadly X" etc. That's where many of those small segments tend to go, instead of going to the dominant ethnicity of the tester, such as "British". Also, in the case that the smoothed segment belonged to another ethnicity you've been assigned, you can view its ethnicity assignment before it was smoothed. Go examine your chromosomes in Gedmatch calcs and see what hilarious results you'd get if every 1cm segment was treated individually.

The reason they're more scared of false positives is because if a customer is assigned an ethnicity they historically cannot have, 23andMe's reputation for accuracy goes down. However, if that same person is assigned a bit more/less of an ethnicity they do have, it isn't going to affect their reputation as much.

[Zoro]

Why do you all default to an agenda when it comes to this shit? It's simply because 23andMe prefers a low false positive rate. That's why they have such high general assignments: "unassigned" , "broadly X" etc. That's where many of those small segments tend to go, instead of going to the dominant ethnicity of the tester, such as "British". Also, in the case that the smoothed segment belonged to another ethnicity you've been assigned, you can view its ethnicity assignment before it was smoothed. Go examine your chromosomes in Gedmatch calcs and see what hilarious results you'd get if every 1cm segment was treated individually.

The reason they're more scared of false positives is because if a customer is assigned an ethnicity they historically cannot have, 23andMe's reputation for accuracy goes down. However, if that same person is assigned a bit more/less of an ethnicity they do have, it isn't going to affect their reputation as much.

Seriously bro. Wtf where’s the politics in this one? I simply agreed with what you just said
but elevates their false negative rate or was my response too complicated for you to get? or was my response too complicated for you?

I just responded to what 23andme says about smoothing segments. I read their article and understood that even large segments are changed as long as they are smaller than the dominant one

There’s nothing wrong with trying to reduce false positives but not when you increase much more false negatives.

I got news for you the ADMIXTURE program which those Gedmatch calculators are based on uses allele frequencies of each INDIVIDUAL SNP nevermind 1cM. 23andme on the other hand uses something like 100SNP haplotype windows

I also agree with what you wrote about the reason for 23andme’s motive but they go way too far when they start altering your larger DNA segments. That’s why only British with very large W. Asian segments from recently such as a W. Asian grandfather or great grandfather get noticeable W. Asian but I think most people would like to know if they have significant non major ancestry even from 500 or 1000 years ago such as would be the case with decent size segment



Also, in the case that the smoothed segment belonged to another ethnicity you've been assigned, you can view its ethnicity assignment before it was smoothed

I have never seen this on their website. Where’s it at?

[Pine]

Seriously bro. Wtf where’s the politics in this one? I simply agreed with what you just said or was my response too complicated for you?

Reread your statement with the emoji. If we have a misunderstanding, then forget about it - it's all good.



I just responded to what 23andme says about smoothing segments. I read their article and understood that even large segments are changed as long as they are smaller than the dominant one

There’s nothing wrong with trying to reduce false positives but not when you increase much more false negatives.

I got news for you the ADMIXTURE program which those Gedmatch calculators are based on uses allele frequencies of each INDIVIDUAL SNP nevermind 1cM. 23andme on the other hand uses something like 100SNP haplotype windows


The difference is that doing what I rhetorically suggested would imply you have a lot of trace regions, none of which would appear in a 23andMe report.



I also agree with what you wrote about the reason for 23andme’s motive but they go way too far when they start altering your larger DNA segments. That’s why only British with very large W. Asian segments from recently such as a W. Asian grandfather or great grandfather get noticeable W. Asian but I think most people would like to know if they have significant non major ancestry even from 500 or 1000 years ago such as would be the case with decent size segment



They would, but their pleasure at discovering such ancestry wouldn't undo the damage caused by those who've been reported clearly false ancestry. And the former group would be skeptical of such a discovery because of the latter group.




I have never seen this on their website. Where’s it at?

Google, my Middle Eastern brother: https://blog.kittycooper.com/2016/12/ancestry-composition-numbers-by-segment-at-23andme-com/