23andMe: Possible fake?
I am going to write the whole post in English so that any person would be able to understand and find it online, not only Russian speakers.
A while ago both of my parents and me bought genotyping kits from 23andMe.com. We were very excited and had a lot of fun discussing the results. However, one observation led me to the conclusion that the data presented in their ancestry analysis is at least sloppy, if not a complete fake. It all started when I discovered that all three of us have markers for Balcan and non-specific Southern European ancestry. The percentage of those markers in my genome was higher than in the genomes of both of my parents, so I decided to take a quick look at which particular loci bearing the markers were inherited by me (using the Chromosome view). To my surprise, the picture that I saw made absolutely no sense.
Below are chromosome views of my father and my mother with highlighted Southern European markers.
And this is my chromosome view with the same highlighted markers.
To summarize, my father and my mother have those markers on chromosomes ## 7, 18 and 19. But for me the Southern European markers are located on chromosomes ## 2, 6 and 17. Now, as a Ph.D. in genetics I do realize the power of recombination. Those markers are long stretches of DNA containing similarities to some reference genomes. They can theoretically disappear in the next generation because that particular piece of a genome was inherited from the other parent. Moreover, they in theory could appear "out of nowhere" when two pieces of DNA that would not by themselves qualify as markers that are significant enough, become significant markers after they recombined. But this cannot happen to all of the markers. I should have inherited at least some of them from my parents and it is very weird that they all developed "de novo" in my genome. This is almost impossible. Now, to prevent any insinuations I should clarify that according to 23andMe.com my Y chromosome was the same as my father's and my mtDNA was the same as my mother's. We also share a significant proportion of our whole genomes and are undoubtedly very close relatives (obviously, a father, a mother and a son). Therefore, what is happening here is a mistake. I checked our Finnish markers afterwards and found similar results, although at least in that case I had some coinciding chromosomes (but not marker locations). So, this mistake is not a unique case and I am sure it is widespread.
I sent a request to 23andMe describing the situation, with pictures similar to those that I showed in this post. Nothing happened. It has been week, and I still have not gotten an answer, although I did get an answer for an unrelated request that was sent later. I consider it an admission that they just do not know how to answer this question.
Personally, I do not believe that all of the data they provide is fake. Most probably, they do sequence those SNPs and analyze them. And I even tend to believe that those percentages that we get in the ancestry results are mostly or somewhat true. I give them a benefit of doubt. But it is clear that the analysis is sloppy. Maybe they do not depict marker locations correctly (this is the best case scenario), or their definition of ancestry markers makes them chaotic enough to appear and disappear randomly. In any case, it is now hard for me to trust these data. And I do not even know what is happening in the medical data. Granted, the analysis in this case is much more straightforward (you either have one SNP variant or another), but there is some anecdotal evidence of gross discrepancies between predicted phenotypes and real ones. I do not think this is genetics fault (our knowledge is awfully incomplete, but lactase enzyme is well known), so it can very well be the result of a sloppy analysis of the existing data.
Given these facts, is it still worth it to pay $99 for something you cannot completely trust? You decide. I still think it was fun and somewhat insightful, however, it now comes to my mind that like many good and new things this one is a little bit premature. Given that the sequencing technology is getting cheaper every year, I am sure that in 10 years we will have a much more comprehensive and thorough way of analyzing our genomes, without newbie bugs and corner cutting, as it obviously is with 23andMe today. On the other hand, that will be in 10 years, and 23andMe is available right now.
A while ago both of my parents and me bought genotyping kits from 23andMe.com. We were very excited and had a lot of fun discussing the results. However, one observation led me to the conclusion that the data presented in their ancestry analysis is at least sloppy, if not a complete fake. It all started when I discovered that all three of us have markers for Balcan and non-specific Southern European ancestry. The percentage of those markers in my genome was higher than in the genomes of both of my parents, so I decided to take a quick look at which particular loci bearing the markers were inherited by me (using the Chromosome view). To my surprise, the picture that I saw made absolutely no sense.
Below are chromosome views of my father and my mother with highlighted Southern European markers.
And this is my chromosome view with the same highlighted markers.
To summarize, my father and my mother have those markers on chromosomes ## 7, 18 and 19. But for me the Southern European markers are located on chromosomes ## 2, 6 and 17. Now, as a Ph.D. in genetics I do realize the power of recombination. Those markers are long stretches of DNA containing similarities to some reference genomes. They can theoretically disappear in the next generation because that particular piece of a genome was inherited from the other parent. Moreover, they in theory could appear "out of nowhere" when two pieces of DNA that would not by themselves qualify as markers that are significant enough, become significant markers after they recombined. But this cannot happen to all of the markers. I should have inherited at least some of them from my parents and it is very weird that they all developed "de novo" in my genome. This is almost impossible. Now, to prevent any insinuations I should clarify that according to 23andMe.com my Y chromosome was the same as my father's and my mtDNA was the same as my mother's. We also share a significant proportion of our whole genomes and are undoubtedly very close relatives (obviously, a father, a mother and a son). Therefore, what is happening here is a mistake. I checked our Finnish markers afterwards and found similar results, although at least in that case I had some coinciding chromosomes (but not marker locations). So, this mistake is not a unique case and I am sure it is widespread.
I sent a request to 23andMe describing the situation, with pictures similar to those that I showed in this post. Nothing happened. It has been week, and I still have not gotten an answer, although I did get an answer for an unrelated request that was sent later. I consider it an admission that they just do not know how to answer this question.
Personally, I do not believe that all of the data they provide is fake. Most probably, they do sequence those SNPs and analyze them. And I even tend to believe that those percentages that we get in the ancestry results are mostly or somewhat true. I give them a benefit of doubt. But it is clear that the analysis is sloppy. Maybe they do not depict marker locations correctly (this is the best case scenario), or their definition of ancestry markers makes them chaotic enough to appear and disappear randomly. In any case, it is now hard for me to trust these data. And I do not even know what is happening in the medical data. Granted, the analysis in this case is much more straightforward (you either have one SNP variant or another), but there is some anecdotal evidence of gross discrepancies between predicted phenotypes and real ones. I do not think this is genetics fault (our knowledge is awfully incomplete, but lactase enzyme is well known), so it can very well be the result of a sloppy analysis of the existing data.
Given these facts, is it still worth it to pay $99 for something you cannot completely trust? You decide. I still think it was fun and somewhat insightful, however, it now comes to my mind that like many good and new things this one is a little bit premature. Given that the sequencing technology is getting cheaper every year, I am sure that in 10 years we will have a much more comprehensive and thorough way of analyzing our genomes, without newbie bugs and corner cutting, as it obviously is with 23andMe today. On the other hand, that will be in 10 years, and 23andMe is available right now.