Results 1 to 7 of 7

Thread: Number of SNPs tested.

  1. #1
    Registered Users
    Posts
    174
    Sex
    Location
    Greece
    Ethnicity
    Hellenic
    Y-DNA (P)
    R-Z2103>R-Y155609
    mtDNA (M)
    W6

    Greece

    Number of SNPs tested.

    According to this comparison, Myheritage tests for 702,442 autosomal SNPS, AncestryDNA for 637,639, FTDNA for 612,272 and 23andme for 630,132.

    So they have the biggest number of SNPs tested out of them all.

    Yet, it seems that on this board (and others) they get a bad rep in relation to any other company.

    What gives? Why this bias?
    mixed mode Oracle for Global25:

    "56.3% Greek + 43.7% Greek_Trabzon"@"0.0192"

  2. The Following User Says Thank You to dosas For This Useful Post:

     Greekscholar (08-11-2019)

  3. #2
    Registered Users
    Posts
    767
    Sex
    Location
    USA
    Nationality
    USA
    Y-DNA (P)
    I-FGC24357
    mtDNA (M)
    H5

    United States of America Canada
    The table is deceptive. Depending on the vendor, maybe a third are not used for genealogical purposes. For the real picture, put the kit on GEDmatch and check the number of SNP's used in comparison and also run it through the diagnostic utility to see how many SNP's GEDmatch can use.

    Before the move to the new chipset, some vendors used most of the SNP's for genealogical testing.

    Things have gone downhill in my estimation with autosomal testing.

  4. The Following User Says Thank You to C J Wyatt III For This Useful Post:

     karwiso (08-15-2019)

  5. #3
    Registered Users
    Posts
    1,216
    Sex
    Location
    Pittsburgh, PA, USA
    Ethnicity
    Ger.-Brit.-Catalan-more
    Nationality
    (U.S.) American
    Y-DNA (P)
    R-YP619*
    mtDNA (M)
    H1bg

    United Kingdom Germany Bayern Catalonia France Ireland Switzerland
    Quote Originally Posted by dosas View Post
    According to this comparison, Myheritage tests for 702,442 autosomal SNPS, AncestryDNA for 637,639, FTDNA for 612,272 and 23andme for 630,132.

    So they have the biggest number of SNPs tested out of them all.

    Yet, it seems that on this board (and others) they get a bad rep in relation to any other company.

    What gives? Why this bias?
    There's a difference between the number of SNPs used for genealogy purposes -- i.e., to determine how much matching you have with another individual -- and the number used for ancestry (ethnicity) purposes. Also, it can be difficult to compare across platforms. Some of the more recent chips actually test for far fewer SNPs than the older ones, though the companies involved might argue that they're more carefully selected. Not all SNPs are equally "ancestry informative".

    For example, 23andMe's v2 chip tested for about 550,000 SNPs and v3 tested for around 900,000. There was some overlap between the two, but if someone tested on both platforms I believe they had over a million tested SNPs. But v4 went back down to 570,000 and v5 is about 640,000. There's even less overlap between v2-v3 and v4 or v5, so only a fraction of the SNPs can actually be compared between the tests. SNPs not in both versions have to simply be ignored.

    At MyHeritage, you're being compared not only to people who tested at MyHeritage but also who tested at another company and had their results uploaded to MyHeritage. So how good the comparison is can depend very much on how much overlap there is between the chips. You're only going to be compared on the SNPs common to both chips.

    For my part, I don't think I have as much of a problem with MyHeritage as some do. I think each testing company has its strengths and weaknesses, and how good its ancestry analysis is for any given customer can depend a lot on luck. By that I mean that different reference panels may just by chance be more appropriate for different people, and you can luck out and find a company that just happens to work very well for you. By the same token, some companies that seem to do well for most people still do not do well for some. Again, I think that may depend more on the companies' various reference panels than on their actual tests.
    Besides British-German-Catalan, ancestry includes smaller amounts of French, Irish, Swiss, Choctaw & another NA tribe, possibly Catawba. Avatar picture is: my father, his father, & his father's father; baby is my eldest brother.

    GB

  6. The Following 5 Users Say Thank You to geebee For This Useful Post:

     C J Wyatt III (08-07-2019),  digital_noise (08-07-2019),  dosas (08-08-2019),  karwiso (08-15-2019),  msmarjoribanks (08-08-2019)

  7. #4
    Registered Users
    Posts
    410
    Sex

    IOW, it's not the size of the thing, it's how you use it. Although I disagree in that "not all SNP's are used for genealogical purposes"; in general all SNP's are used for DNA matching (Admittedly, some medical SNP's are so rare that they contribute less to relative matching, but they do anyway). Why it may feel like they aren't is that the same SNP has to be tested in every sample that is compared, and with increasing number of different tests, this becomes an issue for sites like MyHeritage which allows people to upload any test and match against others.

    As for ancestry estimation, it depends on the algorithm. Many methods make the assumption that the SNP's used must be statistically independent, ie. not inherited together in the population considered. This means that only 200,000 - 300,000 SNP's total can be used, no matter how many are tested. Because the ancestry estimations are concerned with *recent* ancestry (ie. past few hundred years), using denser SNP's would not automatically provide relevant information.

    Of course, the original post rather prompts the question of bad rep by whom, and for what reason. Answers to those questions will probably provide better answer to the original question...

  8. The Following User Says Thank You to Donwulff For This Useful Post:

     DMXX (08-07-2019)

  9. #5
    Registered Users
    Posts
    1,216
    Sex
    Location
    Pittsburgh, PA, USA
    Ethnicity
    Ger.-Brit.-Catalan-more
    Nationality
    (U.S.) American
    Y-DNA (P)
    R-YP619*
    mtDNA (M)
    H1bg

    United Kingdom Germany Bayern Catalonia France Ireland Switzerland
    Quote Originally Posted by Donwulff View Post
    IOW, it's not the size of the thing, it's how you use it. Although I disagree in that "not all SNP's are used for genealogical purposes"; in general all SNP's are used for DNA matching (Admittedly, some medical SNP's are so rare that they contribute less to relative matching, but they do anyway). Why it may feel like they aren't is that the same SNP has to be tested in every sample that is compared, and with increasing number of different tests, this becomes an issue for sites like MyHeritage which allows people to upload any test and match against others.

    As for ancestry estimation, it depends on the algorithm. Many methods make the assumption that the SNP's used must be statistically independent, ie. not inherited together in the population considered. This means that only 200,000 - 300,000 SNP's total can be used, no matter how many are tested. Because the ancestry estimations are concerned with *recent* ancestry (ie. past few hundred years), using denser SNP's would not automatically provide relevant information.

    Of course, the original post rather prompts the question of bad rep by whom, and for what reason. Answers to those questions will probably provide better answer to the original question...
    Well, yes and no. I wasn't actually agreeing with the idea that "not all SNPs are used for genealogical purposes". I'm not sure that I would consider an "ethnicity estimate" to be a "genealogical purpose".

    To me, "genealogy" is the study of one's own lineage. Ancestry is obviously something you're looking at, but from the standpoint of trying to determine who your ancestors actually were -- not simply where they might have come from.

    Except for 23andMe, I don't know of any company that gives the possibility of being able to correlate ancestry (in the sense of "ethnic origin") to your actual family tree. Because Ancestry Composition is presented in the same form as match comparisons, you can sometimes get an idea of how a particular ancestry is associated with particular ancestors.

    For example, I have a 3rd cousin/3rd cousin once removed who is related to me through my maternal grandmother's maternal grandparents. I don't put the relationship the way I do because I'm uncertain, but because he's actually both of these. His father is my 2nd cousin once removed, and his mother is also my 3rd cousin -- so their son is both my 3rd cousin and my 3rd cousin once removed.

    Anyway, the two of us share a Native American ancestor from several generations ago, and one of the segments the two of us share is in a region identified as Native American for both of us. My maternal grandmother's maternal grandparents -- who were 2nd cousins (yes, my tree here is very tangled) -- both descend from this ancestor. This isn't proof that this particular segment came from her, but it's certainly possible.

    With most companies, this kind of comparison isn't possible. Just percentages.

    So, from my standpoint, most companies generally do use as many SNPs as they can for genealogy. They don't necessarily use them for determining "ethnicity", which as I said is not something I consider to be genealogy. (Even if it is connected to genealogy.)

    And here you're right about size. Not all SNPs are equally "ancestry informative". Some SNPs are widely spread across different populations, for example. This is one of the arguments some companies have used to say that their more recent chips are actually better at determining "ancestry" (in the sense of where your ancestors came from) than older chips. But when it comes to determining DNA matching, more SNPs are better -- or rather, more overlapping SNPs are better. So comparing on the same chip version, for example, would usually be preferable no matter what version it is.

    Supposedly, when it comes to SNPs with medical significance, some companies actually prefer to avoid them since they don't want their product to be deemed a "medical device" by the FDA. 23andMe is really more of an exception, though I can't say whether they'll remain that way.

    But, again, as you say it isn't just "size" but "selection". Or so the companies that do the testing like to say. In my own opinion, what matters most is actually the quality of the reference populations. However, even here you don't want to just compare SNP-by-SNP. You actually want to use haplotypes, and potentially if you have greater SNP density you might be able to use somewhat shorter blocks.
    Besides British-German-Catalan, ancestry includes smaller amounts of French, Irish, Swiss, Choctaw & another NA tribe, possibly Catawba. Avatar picture is: my father, his father, & his father's father; baby is my eldest brother.

    GB

  10. The Following 2 Users Say Thank You to geebee For This Useful Post:

     dosas (08-08-2019),  karwiso (08-15-2019)

  11. #6
    Registered Users
    Posts
    385
    Ethnicity
    Greek
    Nationality
    American

    United States of America Greece
    Quote Originally Posted by dosas View Post
    According to this comparison, Myheritage tests for 702,442 autosomal SNPS, AncestryDNA for 637,639, FTDNA for 612,272 and 23andme for 630,132.

    So they have the biggest number of SNPs tested out of them all.

    Yet, it seems that on this board (and others) they get a bad rep in relation to any other company.

    What gives? Why this bias?
    I really like MyHeritage for Greeks and neighboring populations. Greeks in my experience, are far more likely to get a majority of their ancestry assigned to their "Greek and Southern Italian" region than with say, Ancestry.com (whose Greece region is very Balkan-centric and makes no formal mention that Southern Italy/Sicily/Island Greeks are part of the same genetic population)and being popular in Europe means lots of cousin matches. The smaller population estimates normally make sense too. Middle East and West Asia for Island Greeks, Balkan for Mainland Greeks.

    MyHeritage seems to lose their bearings sometimes with people with ancestors from many different places. I have some cousins who are 1/2 Greek or less, and get nothing in the Greek region, but instead retain the minor regions like West Asia or the Middle East instead.

  12. The Following User Says Thank You to Greekscholar For This Useful Post:

     dosas (08-12-2019)

  13. #7
    Registered Users
    Posts
    15
    Sex

    Quote Originally Posted by C J Wyatt III View Post
    The table is deceptive. Depending on the vendor, maybe a third are not used for genealogical purposes. For the real picture, put the kit on GEDmatch and check the number of SNP's used in comparison and also run it through the diagnostic utility to see how many SNP's GEDmatch can use.

    Before the move to the new chipset, some vendors used most of the SNP's for genealogical testing.

    Things have gone downhill in my estimation with autosomal testing.
    Quote Originally Posted by geebee View Post
    There's a difference between the number of SNPs used for genealogy purposes -- i.e., to determine how much matching you have with another individual -- and the number used for ancestry (ethnicity) purposes. Also, it can be difficult to compare across platforms. Some of the more recent chips actually test for far fewer SNPs than the older ones, though the companies involved might argue that they're more carefully selected. Not all SNPs are equally "ancestry informative".

    For example, 23andMe's v2 chip tested for about 550,000 SNPs and v3 tested for around 900,000. There was some overlap between the two, but if someone tested on both platforms I believe they had over a million tested SNPs. But v4 went back down to 570,000 and v5 is about 640,000. There's even less overlap between v2-v3 and v4 or v5, so only a fraction of the SNPs can actually be compared between the tests. SNPs not in both versions have to simply be ignored.

    At MyHeritage, you're being compared not only to people who tested at MyHeritage but also who tested at another company and had their results uploaded to MyHeritage. So how good the comparison is can depend very much on how much overlap there is between the chips. You're only going to be compared on the SNPs common to both chips.

    For my part, I don't think I have as much of a problem with MyHeritage as some do. I think each testing company has its strengths and weaknesses, and how good its ancestry analysis is for any given customer can depend a lot on luck. By that I mean that different reference panels may just by chance be more appropriate for different people, and you can luck out and find a company that just happens to work very well for you. By the same token, some companies that seem to do well for most people still do not do well for some. Again, I think that may depend more on the companies' various reference panels than on their actual tests.
    I tested with MyHeritage, and the results in GEDmatch seems to be very good and they are on pair with my ancestry. My sister tested with FTDNA, and some of her results are a little weird in GEDmatch. I know siblings are not supposed to score the same, but the difference is a little weird. I'm exactly 1/32 Amerindian in real life, and I score around ~3% Amerindian in GEDmatch. So it totally makes sense. But my sister scores 5%. Is my father 10% or 6% Native American after all? It's a big difference, and the second option makes more sense. What raw data should I trust? I'm not an expert, but since MyHeritage tests with more SNPs, maybe their raw data it's better? Thanks in advance!

  14. The Following User Says Thank You to Genetic For This Useful Post:

     C J Wyatt III (10-02-2019)

Similar Threads

  1. Genetic Company, Number of SNPs used and Reliability
    By Lupus82 in forum Inquiries Corner
    Replies: 0
    Last Post: 02-16-2018, 10:23 PM
  2. Number of Matches at 111 Markers
    By GoldenHind in forum Other
    Replies: 5
    Last Post: 10-17-2017, 06:03 PM
  3. What YDNA SNPs are tested at Livingdna?
    By RedBlazedNugguts in forum Living DNA
    Replies: 7
    Last Post: 08-29-2017, 05:38 AM
  4. List of Y-snps tested by 23andme chip
    By Afshar in forum 23andMe
    Replies: 7
    Last Post: 08-20-2015, 12:07 PM
  5. What's up with Chromosome number 6
    By vettor in forum Autosomal (auDNA)
    Replies: 2
    Last Post: 11-18-2014, 07:07 PM

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •