Why doesn't 23andMe test entry propagate the mtDNA & yDNA haplogroup?

+11 votes
375 views
The 23andMe autosomal test entry allows the addition of Mitochondrial haplogroup and Y-chromosome haplogroup, but filling these in does not result in the mtDNA going up the mother's line, nor the yDNA going up the paternal line like is done with a yDNA or mtDNA test from FamilyTreeDNA.com.

Can we get this info to propagate up the tree for yDNA and mtDNA?
in Genealogy Help by William Foster G2G6 Mach 7 (72.1k points)
retagged by Ellen Smith
I now believe the benefits of having (23andMe, LivingDNA and Genographic Project reported) haplogroups propagate up their respective direct lines outweighs the problems that may arise.  This is because we don't currently have public databases for mtDNA HVR1&2 differences or Y chromosome haplotypes.

I sincerely hope WikiTree will make it possible for those haplogroups to be automatically associated with their respective direct lines.

I'm not entirely certain I follow, William. Since the 23andMe and LivingDNA tests are not for-purpose for yDNA and mtDNA (they're only a by-product of the one-run autosomal test on the customized GSA microarray chip) and can only present limited haplogroup information, would we really want that detail propagating beyond the current parameter of eight generations already set for autosomal results? There would be no matching, comparisons, or reference-sequence differences, only a haplogroup...which might well differ from the results provided from actual yDNA/mtDNA testing (Roberta Estes wrote about this just last Wednesday).

Member-entered, the 23andMe yDNA and mtDNA haplogroups are displayed with the test information on the member's profile. I could see expanding the test description thumbnails under the "DNA Connections" panel on the right-hand side of profiles to include the haplogroups with the auDNA information. But it wouldn't seem to make sense to treat 23andMe haplogroup estimates as actual yDNA or mtDNA tests and propagate them as such through the tree without the auDNA generational limits. We have over 19,000 23andMe test takers in the tree, and a high-level haplogroup--by itself--often has little genealogical relevance. Viewing the profile of a male born, say, 1600 and seeing a hundred "DNA Connections" by dint of autosomal results that returned a haplogroup of R-M269 would be, I think, a very daunting list to sift through; would be confusing to those, and there are many, who don't understand that "DNA Connections" does not mean "DNA Matches"; and would offer no other value because no further comparisons between test takers could be made.

I agree with both sides of the issue (unfortunately)-;

If Y haplogroup info from 23andMe, LivingDNA, Genographic, etc. was automatically associated with direct paternal lines, then there would be cases when two direct paternal line cousins would discover they MIGHT be a match if they took the next step by testing their Y STRs to see if they had sufficiently matching haplotypes.  I don’t believe WikiTree should allow “confirmed with DNA” in this situation unless there were sufficiently matching haplotypes or matching terminal SNPs (“next generation” Y testing) which have been demonstrated to be in a genealogical timeframe.

Similar for mtDNA (sufficiently matching HVR differences).

Sincerely, Peter
Edison, very very respectfully because this may be the first time(!), I have to disagree with you.  I have extensive FTDNA yDNA results and I took the Living DNA test (for fun and verification), and I was absolutely amazed at the quality of their haplogroup result.  It was apparently designed with Y chromosome SNP's based on the YFull tree, and provided a haplogroup down to the last SNP they have in my haplogroup trail, R-L257.  That is well past the M269/M343 SNP bundle pack and well into the Z18 pack.  (And more amazingly, it was just one test of the 3 they bundle for about $99, so for about $33 I got the equivalent of what costs hundreds on FTDNA.   But of course does not include the very useful STR's.)

If we were to apply a quality limitation on yDNA tests, then you would have to consider banning all STR predicted results instead, which would probably eliminate most of the current yDNA haplogroup info on WikiTree.  I can't speak for 23andMe, but the Living DNA result is greatly superior to the STR high level predicted haplogroup results.  Mine is only one data point, but for me it provided a haplogroup with 'terminal' SNP to the FULL extent that YFull could determine, just 2 branches short of FTDNA.
When it comes to an apparent haplogroup conflict on an older gentleman's profile, we might consider selecting the best one, overriding lesser ones, so long as they are roughly equivalent, only different in precision.  STR-based predictions would automatically be eclipsed then.  If they conflict in substance, then both should be displayed, as that would indicate a serious problem in one of the descendant lineages.

Rob, heya!  laugh I know more people who disagree with me than agree, so you're you're in a very popular club!

I dashed that last post off with a 7:00 a.m. deadline looming this morning. That's why it was so short <cough, cough> and, reading it back, I wasn't terribly clear. See Ellen's answer below: the primary issue to me isn't one of test quality (though you just know I'll have a comment about that) but one of test functionality.

The assumption I made, which I still do, is that this is about getting information from individual test takers to propagate throughout the tree and be listed as discrete clusters of links under the "DNA Connections" panel on ancestors' profiles. Unfortunately, far too many people think that actually means "DNA Matches," but that aside I feel the intent of the feature--at least my interpretation--is to have a way for WikiTreers to find others, descended from the same ancestor, who have taken a DNA test with which they can compare. The issue for me is that a haplogroup-only designation from 23andMe or LivingDNA isn't something that offers any capability of comparison. To my knowledge, neither company has any plans to offer comparisons and matching for yDNA or mtDNA. They would almost certainly have to augment their testing procedures to do so.

Knowing the presumed haplogroup of an ancestor definitely has value. But if the 23andMe autosomal DNA entry is shown on the profile--and that propagates back to 6g-grandparents--one click on a person's "test details" under that "DNA Connections" panel shows me both the mtDNA and yDNA haplogroups of the test taker, if entered. And I can also see if there is any matching that might be done at 23andMe or GEDmatch.

A haplogroup is a static piece of information. There are no matchings or comparisons to be done unless a testing service provides the tools or services to do so. (Okay; arguably, you and I could exchange our raw data files from 23andMe, plug them into a spreadsheet with about 700,000 rows, identify all the pertinent reference clusters, and devise a comparison method that way. But, realistically...) My example about the guy born in 1600: if the PM noted the haplogroup(s) on the profile, it would serve the same purpose as having a hundred different automatic entries under "DNA Connections" that serve no actionable purpose because it's too far back for autosomal DNA to be of use, and there's no yDNA/mtDNA comparison available.

(BTW, Peter and I have discussed this in the past, and Ysearch and Mitosearch served, IMHO, a very real need in the genetic genealogy landscape. That they stopped being maintained about 10 years ago was a massive migraine, but their functionality is absolutely needed. I know of no takers on the horizon, though. For many months I'd hoped that Curtis Rogers might take up the challenge and include the services in GEDmatch, but I believe Curtis is...er, plenty busy with GEDmatch as it is. Don't think that's gonna happen. FWIW, the Ysearch and Mitosearch closures made news, but we also lost another a few months before: "mtDNA Community," mtDNAcommunity.org, shut down on January 5.)

Okay. That was my point. Not about test quality, but about having a hundred different WikiTreers showing up under an ancestor's "DNA Connections" panel with only a haplogroup designation and nothing actionable. That's what doesn't seem useful to me.

Now, the tests. Which really is a different subject. And I'm not even talking specifically about this WikiTree DNA FAQ entry. Note: this next bit is a dive down an only semi-related rabbit hole. My only real point was above. About automatically propagating tens of thousands of non-actionable haplogroup designations through the tree. So this part is a walk into deep weeds. Stop reading now. Seriously. You have been warned. Stop reading!

And...I ran afoul of the 12,000-character limit again. I mean, who can write anything in less than 12,000 characters? Are you surprised I'm not a big Twitter user? So...

Part 2

Rob, your LivingDNA results were not from three tests bundled together for one price. They just weren't, despite LivingDNA's marketing claims. Neither LivingDNA nor 23andMe run for-purpose yDNA or mtDNA tests. Here's what happens.

They buy the Illumina Infinium Global Screening Array bead-chips in bulk. That's the only chipset they and 23andMe currently use for genealogy/population testing (FTDNA, Ancestry, and MyHeritage are still using the OmniExpress chip, but that may change in the future). Out of the box, the GSA chip looks at around 640,000 autosomal markers (including just over 17,000 on the X-chromosome), 1,456 yDNA SNPs, and 137 mtDNA SNPs. The microarrays can be configured (relatively easily, but it is a process that has to be done with all chips before testing begins) for up to an additional max of <50,000 company-specified target loci.

We don't know how 23andMe and LivingDNA customize their GSA chips. Somebody could gather a number of sets of raw data results and probably figure it out, but I would imagine that they make tweaks from time to time to the programmed chip configuration.

It's a minimum three-day lab process to get the results. On day one is an amplification and incubation procedure that has to sit overnight. On day two that incubated sample goes through enzymatic fragmentation, after which is a stopping point if the lab decides to batch and resume later or the next day. The fragmented DNA then undergoes an alcohol precipitation procedure (another possible stopping point). Then the sample material goes back into the suspension material. The next step is called hybridization; this is where the suspension is washed over the (now presumably customized) microarray chip and the fragmented DNA strings align to the beads on the chip...the remarkable technology where the microarray is able to trick the DNA into thinking it is attaching to its complementary nucleotides, adenine to thymine, and cytosine to guanine. The suspension is left on the chip overnight. On day three (or four or five), there's another enzymatic procedure and then a fluorescent stain is applied. The final step--other than any analyses performed--is to actually image the microarray chip and convert all the squishy biological stuff to digital data. The actual imaging step only takes about three minutes.

Pretty fascinating stuff. The chip itself can be reused a couple of times once ablated and sterilized. Bottom line is that this is the only test 23andMe and LivingDNA perform. They do not do separate yDNA or mtDNA testing; they do not do multiple passes on selected SNPs. Wouldn't be remotely feasible, economically, to do so. If the entire process goes smoothly and the overall results from imaging meet minimum overall quality control standards, away you go.

We all have no-calls in our raw autosomal data, and there will be no-calls, almost certainly, in the targeted yDNA and mtDNA SNPs. For example, your Z19 and Z14 Y-SNPs may not even have been tested, and S375 may have been a no-call, but since L257 showed up as positive, L257 you became. And looking at the lists of WikiTree 23andMe test takers the Y-haplogroups entered are all over the place; I have no idea how granular they may have recently become, but the haplogroups on WikiTree range from I1 to I-Y7232 and R1 to R-PF6570. I have to think that some people who also tested at FTDNA included their deepest-known SNP in their 23andMe entry even if that wasn't the 23andMe reported result.

Again, I don't know how 23andMe and LivingDNA customize their GSA chips for those additional <50K markers available, but I think it's a pretty good bet that 23andMe, at least, don't focus the majority of their targets at the Y-chromosome or mtDNA molecule. They're no doubt programming the chips for more medically-relevant, gene-specific loci. Can't comment about LivingDNA. But remember also that the marketing wars are all about ethnicity and admixture, and the GSA chip is multi-purpose; that isn't its main goal. If I had to guess, it would be that the companies are using a sizable portion of their customizable beads to match loci with whatever population genomics database(s) they're using.

But even if LivingDNA decided to expend all their chip customization on yDNA, they fall far short of known-stable SNPs today, and the counts are growing regularly as more and more Y-chromosome full-sequence tests are completed and evaluated. The estimate right now is that about 385,000 stable yDNA SNPs have been identified; the ISOGG Y-haplotree alone has 76,352 cataloged; YSEQ offers individual tests for 80,659.

Oh, an aside: Yfull isn't actually "full" by any stretch; they catalog as tests are sent to them to process. For example, this is as far down as Yfull goes for my Williams clan, compared to the additional 40 or so SNPs that have been cataloged by Alex Williamson at The Big Tree (which, of course, doesn't include novels or InDels, insertions/deletions).

FTDNA and YSEQ offer for-purpose yDNA (STR and SNP) and mtDNA testing. I really can't comment about Oxford Ancestors: Bryan Sykes announced their closure last March, and then in May said that they would remain open. But the pricing and services are...odd...as in US$265 for a 26 Y-STR test or a 400-SNP mtDNA test (HVR1 alone is 569 base pairs, and HVR2 is 575 base pairs, so I'm not sure what they're testing). 

When FTDNA and YSEQ do a SNP panel for yDNA or mtDNA, they look at every SNP defined by the panel or region, not selective loci set by an autosomal microarray chip. My bet is that if you check your Y-SNP results from FTDNA, every loci tested will be shown, and each will show as having tested with a positive or negative result. There are, as a matter of course, no-calls in the full-sequence test results, but a no call in a SNP panel is exceedingly rare (I haven't actually seen one) in part because the companies run multiple passes on each loci, not one-and-done like results from the GSA chip. Likewise, Y-STRs are fluoresced multiple times so that the amplitudes--representing the allele repeats--can be examined in detail and a determination made as to the correct, or modal value.

With mtDNA, we all carry more than one mitochondrial haplotype within us--but seldom more than two, and those differences are rarely more than a single base pair, not counting InDel copies--sometimes within a single one of our cells. The only way mtDNA changes is via heteroplasmic mutation. You have over 600 trillion (yep; with a "t") mitochondria in your body at any given time, happily working to produce ATP and to reproduce themselves. All plants and animals love their mitochondria. Even at the glacial rate of general mtDNA mutation (call it about 0.48 bp/My, per base pair per million years), copy errors and oddities happen.

The GSA chip looks only at the specifically programmed loci. It will miss any InDels (the result will be a no-call for a deletion and insertions will simply be overlooked), and there can be no checking to verify whether the reported allele at a particular locus is the result of a heteroplasmy or represents the mtDNA genome that's present in the other 99.9999% of those 600 trillion mitochondria. For a heteroplasmy to "stick" and start becoming the dominant mtDNA genotype, or homoplasmy, requires many, many generations. I can't talk to YSEQ's mtDNA testing, but since FTDNA does multiple passes for each loci, they have a whole nomenclature developed to describe detected heteroplamic variations and have established a threshold of 20% for such variations: if the multiple passes indicate a concentration lower than 20% of the less-common allele, it will be considered anomalous, not a heteroplasmy worth measuring, and they'll move on.

Told ya it was a rabbit hole, and I did warn you to stay out of it and stop reading. But there is, with the current technology and techniques, very much both a qualitative and functional difference among tests. There are good reasons that specific tests exist for yDNA and mtDNA.

Hm. Maybe doubling the 12,000-character post limit? I think I can work within 24,000-25,000.  angel

Thank you, Professor Williams, really great class!  And you can consider me signed up for your next classes too!  I learned quite a bit!   smiley

However ... I still think other yDNA results should be propagated.  I'm sure they must exist, but I don't think I have ever seen a profile with more than 1 yDNA result, and if I did, it would be very interesting.  I would immediately be looking for conflicts.  I know I've seen profiles with 20 or 30 autosomal results, and they looked fine.

Furthermore, if you're going to turn up your nose at my Living DNA result (based on multiple SNP's at least), cheeky  how can you possibly stand all the STR-based absurdly broad and somewhat unreliable predicted haplogroups !?!   wink

Dear Professor Jacobson:

Many, many thanks for your letter of 6 July, professor. But I did warn you more than once not to read the final 1,500 words...

As with Ellen's answer, I have no problem at all if people want to manually enter an "Other mtDNA" or "Other yDNA" test. After all, you can take a deep dive using YSEQ in Germany, as an example, but it isn't available as a testing company choice in profiles, I imagine largely because of market share and lack of matching/comparison tools. And if someone wants to enter their 23andMe or LivingDNA haplogroups that way, more power to 'em. That information may have a specific bearing on their research or on a particular ancestral line.

Thing is, though, 23andMe provides an mtDNA haplogroup with every set of results, and a yDNA haplogroup if applicable. Let's see...WikiTree currently shows that 6,644 members have posted that they've taken an FTDNA yDNA test (of any kind); 5,727 who have taken an FTDNA mtDNA test (of any kind); and 19,183 who have taken a 23andMe test. Not all of those will have a haplo listed, but most do. So if somebody flips a switch to automatically propagate 23andMe haplogroups as separate yDNA and mtDNA tests under "DNA Connections," overnight we would at least triple the number of mtDNA test entries and at least double the number of yDNA tests propagating throughout the tree onto many thousands of profiles, entries that show only a haplogroup designation and that offer no possibilities of genealogically-relevant comparison or matching. Too, because those entries won't stop at eight generations, they'll populate as far back as a profile is connected on the tree. Not Armageddon, but not terribly useful.

I've seen multiple Y tests shown on profiles, but have no idea how to go about searching for that. Never many, though, on a single profile. If I could talk cousins into joining WikiTree, there would be four of us on my Williams line, which still brickwalls at 1790. There are 14 of us we know share a paternal MRCA almost certainly no earlier than the mid- to late-1600s, but there are five branches there that are yet to be connected by paper-trail.

"...How can you possibly stand all the STR-based absurdly broad and somewhat unreliable predicted haplogroups !?!"

Ya know, a lot of people point to Jim Bartlett when they talk about autosomal triangulation (although sometimes misreading his comments). When folks ask him about how he organizes people into triangulation groups, he'll quickly correct them and say that his triangulation groups don't contain people. He doesn't triangulate people: he triangulates segments.

I pretty much view haplogroups the same way. Interesting anthropologically, and useful to invalidate a hypothesized ancestral relationship, but otherwise haplogroups aren't worth much for genealogy until you're deep into the terminal Y-SNPs that can only be uncovered with full-sequence testing, and not at all for mtDNA. For my line, we're down to BY22194--over 50 SNPs deeper than M269--before there's any reasonable time/geography correlation, and the current thinking is that puts an MRCA sometime around the middle of the first millennium through around the start of the second. But at least we're Anno Domini. I've seen people--just this week, in fact--try to claim a Y-SNP that bifurcated 4,600-3,900 YBP is genealogical proof of ancestry.

The high-level STR predictions of haplogroup are generally fairly good. At least you'll never see someone who's an E predicted to be an R. In the FTDNA projects, though, if not confirmed by SNP testing the Y-haplogroup is displayed in red text to show it's only a guesstimate. They even flag the "Niall of the Nine Hostages" badge based on STR results only, but it's all kinda back to people vs. segments: Y-STRs are great for estimating generational ranges in recent genealogies, and not so great for anthropology; Y-SNPs are great for refining date ranges in anthropological and early-surname timeframes, but not so great for recent genealogy (discounting, of course, finding a couple of cousins who want to spend $800 each to see novel SNPs and find out essentially the same thing cheaper STR testing would tell them).

Now, those distinctions will, I believe, begin to blur during the coming years. HiSeq testing techniques are only--from a practical standpoint--a few years old, and we're just starting to see peer-reviewed anthropological and population genetics studies published that really dive into some DNA density in those ancient remains. Pretty excited to see what the next few years bring.

BTW, the "Niall of the Nine Hostages" ancestry is (currently) determined by 20 STRs. The same person, above, who claims the 4,600-year-old SNP as proof of genealogical ancestry also claims descent from Niall...but at those 20 STRs he's a genetic distance of a whopping 10. Some enterprising "Indiana Jones" genetic anthropologist should go find ol' Niall, dig him up, and do a right HiSeq NGS service for him.

And I didn't dis' your LivingDNA test! I dis' pretty much all the major testing companies for their marketing spiels (e.g. "Twice the detail of other ancestry tests," "tests five times more regions than other DNA tests," "3-in-1 ancestry test") but I have no doubt that LivingDNA's tests are excellent for what they can offer. I'll probably hold off until they begin the planned matching services, but they're on my to-buy list. Considering the number of UK-based DNA testing companies that have gone out of business, and that a ton of us here in the New World trace back to the UK, I most genuinely and sincerely want a UK company to grow and thrive.

Yours truly,

E. Williams, Bl.D.

cheeky

Quite Excellent Edison!

Just as a funny aside to offer re testing: I am and have always tested with FT, and continue to do so. BUT, a number of years back, some of us in the very early testing process at FT experienced "kit mixup". This meant that a few 700 level kits were misidentified, "mixed up",  and improperly placed in results. Basically mislabeled results. Well, we spent years trying to connect to cousin Joe out of SC into MS only to find that he actually had Gary out of NC results posted. Change of haplogroups. WOW!!! So we gained cousin Gary and are not cousin Joe's Nimrod in Attala SC! What an amazing correction that made in terms of Mitchell research for our line! Bonus, we all got some free testing c/o Bennett.

If I may offer the tweet version of one of Edison's* many salient points:

FTDNA has specific tests for yDNA and mtDNA. LivingDNA, as far as we can tell, interpolates them with data from their auDNA tests.

So, I could see putting people in categories based on haplogroups and higher-level subclades, but there isn't much point in propagating something huge like E-V68. 

*You do you, Edison--I'm just poking fun at your verbosity ;)

2 Answers

+7 votes
 
Best answer

For years this has been discussed here as an issue of test quality, but I believe the real answer to William's question is that the DNA-propagation functionality in WikiTree is programmed to handle only one set of results per test. So to get our 23andMe mitochondrial and Y-DNA results to propagate, we need to enter them as "Other mtDNA" and "Other yDNA."

Maybe WikiTree would consider adding "23andMe yDNA" and "23andMe mtDNA" as separate test types.

by Ellen Smith G2G6 Pilot (958k points)
selected by William Foster
I suggested the same to a test taker, but wonder if that is acceptable to our standards?
The use of the "Other test" data fields is advice I received a long time ago.

yes If someone specifically wants to add an "other test" I personally see no reason not to. That could reflect anything from forensic/legal autosomal STR tests, to someone who may have taken single Y-SNP tests from YSEQ, to Genographic 2.0 tests that have no place else to record them.

Oh, and Ellen? Don't even look at what I'm about to post...

+2 votes
You can bring your 23 and Me autosomal over to the familytreedna and then order the YDNA and mtDNA test there! It is a more complex test. WE FOUND THE ANCESTORS WE NEEDED to help us go back over 100 years!! We are happy we did it! :)
by Mary Tyler G2G4 (4.8k points)

Related questions

+3 votes
1 answer
+4 votes
2 answers
80 views asked Apr 19, 2018 in WikiTree Help by Mark Dorney G2G6 Mach 2 (24.4k points)
+11 votes
2 answers
+4 votes
1 answer
136 views asked Mar 2 in The Tree House by Karen Raichle G2G6 Mach 6 (64.5k points)
+8 votes
0 answers
100 views asked Nov 6, 2018 in The Tree House by Rob Jacobson G2G6 Pilot (120k points)

WikiTree  ~  About  ~  Help Help  ~  Search Person Search  ~  Surname:

disclaimer - terms - copyright

...