Which DNA test should my father take?

+3 votes
248 views
I recently got my DNA results through 23andme. I chose them because I heard you could upload them to GEDMATCH, FTDNA, etc. Now that I have the results I have found out some of the limitations because the 23andme test was version 5. The results can't be uploaded into the original GEDMATCH database, limiting the pool of potential matches to those in GEDMATCH Genesis. I also can't seem to upload the results to FTDNA. (The FTDNA website says you can upload 23andme V3 and V4. I presume that my problem is that mine is V5.)

Now my father is interested in getting his DNA tested and I don't know which company to go with for him. We will want to compare with each other but we want a broad pool of people to compare with. Should he go with Ancestry and will just have to bite the bullet and pay for another test through Ancestry as well?

Suggestions?
in The Tree House by Nancy Reid G2G3 (3.8k points)

3 Answers

+2 votes
 
Best answer
This is very simple (IMHO) . . .

For yDNA and mtDNA: test at FTDNA.

For auDNA: Since 23andMe and Ancestry do not allow uploads from other providers, one should test at the former only to be in one more pool and/or for health information but should certainly test at Ancestry to take advantage of their large pool.  Results from Ancestry (and 23andMe if testing both) can be uploaded to GEDmatch, MyHeritage, FTDNA, and LivingDNA all for free.  N.B. there might be some advantage to testing at LivingDNA instead of uploading results if there is significant British heritage.
by Thom Anderson G2G6 Mach 3 (38.4k points)
selected by Nancy Reid

Just a quick note. Tests taken at 23andMe after August 2017 cannot be uploaded to FTDNA; that's when 23andMe switched to the GSA chip.

Living DNA can accept any other company's format, but there are caveats: at present, they still do not offer matching/comparing services; uploads go into the One Family Project and there is no ethnicity/admixture breakdown provided; and even though raw data from some other testing companies will include some yDNA and mtDNA SNPs, Living DNA won't attempt to read those data because the reference sequences tested won't match exactly to the ones they test with their customizations to the GSA chip.

OpEd only: the merging of GSA and OmniExpress BeadChip data is very new. I believe the jury is still out as to how accurately this can, as yet, be done given that only 20%-23% of the SNPs tested are the same for both chipsets. There's a whole bunch of genotyping imputation at the heart of any attempt to meaningfully compare results between the two chipsets, and the possibility exists of imputation-caused errors and omissions. I think time and some experiential evidence will give us a better picture of the state of the technology.

+6 votes
A lot depends on what you want to do with DNA research. If you want to also participate in DNA surname study, FamilyTreeDNA.com is where to get Y DNA tested. They also have autosomal DNA testing and their results can be uploaded to GEDMATCH. If all you care about is autosomal DNA then Ancestry has the largest database but the FTDNA has gotten me more matches of people who care about things other than ethnicity. Since you can upload autosomal from both Ancestry and FTDNA to GEDMATCH it doesn't matter so much. I've done both companies and also have Y and mitochondrial tests along with autosomal with FTDNA.

So, it depends on what you want to do.
by Doug McCallum G2G6 Pilot (341k points)
Meant to mention that FTDNA can add tests and upgrades without having to send another sample.
Echoing what Doug said, a differentiator for FTDNA is that they store the submitted sample and it's then available for future tests. This can be an important factor when choosing where to test older family members. I had a sample remain viable with FTDNA through over a dozen test upgrades from 2003 through last April. The sample should last around 25 years if not depleted from testing (do take care to follow the sampling instructions: the quality and longevity of the sample is all about thorough cheek swabbing).

And basic Y-chromosome testing would not deplete the sample significantly. You could buy the autosomal test to start, and then take your time deciding whether yDNA or mtDNA testing is of interest. Even upgrade a basic Y-37 test to 111 markers later, as well. AncestryDNA has by far the largest data pool of autosomal DNA tests, but they don't do yDNA or mtDNA, and they don't store the sample for future use.

The caveat here is that newer tests--Next Generation Sequencing for testing the full Y-chromosome, the complete mitochondrial genome, etc.--require more initial sample material, and these are the kinds of tests the future holds for DNA testing for genealogy. Moderate resolution, full-genome testing is already available for a price-point around or just above $1,000, down by a five-fold cost factor in just a couple of years. But even if direct-to-consumer full-sequence tests become popular well before I thought they'd be, they will probably always require a fresh, discrete sample.
One thing I'd like to do is identify any relatives from my mother's side. She was adopted and we have absolutely zero information about her bio parents. One reason I did the 23andme test was that it 'included mtDNA' haplogroup. I thought this might help. But it seems not, as the results aren't very specific and seem of little use. Once my father tests, I can at least find out who my connections are who aren't related to him, which will help me figure out who is related to me on my maternal side.
Well, if you're looking for your mother's relatives, then your father's test won't be of much help -- unless your father and mother are related to each other
autosomal testing would be the most effective way to find your mother's side. Doesn't really matter which testing service for that but as has been said, Ancestry has the biggest customer base. The downside I found was the lack of interest from people who were close matches. No responses from messages sent to most. I got a better response rate at FamilyTreeDNA. If you are only doing one test, Ancestry is probably best since you can download from them and upload to the others. I've had more engagement from FTDNA users. MyHeritage (mentioned below) has a growing base. mtDNA is not really useful for finding near relatives. I've only had 6 matches at FTDNA for that.

Do you have any siblings? Siblings can sometimes match different relatives since you don't get all of your mother's DNA. One DNA expert told me that 3 siblings will get a good picture of the grandparents. Having your father's will help eliminate his side. Chromosome mapping/painting will help. You want to know which markers come from your mother. Doesn't really matter who's testing for that.

Nancy: Yep; you're on the right track. There are nifty tools that can help you keep track of the effort to align segments from your autosomal matches with your paternal and maternal lines...and their paternal and maternal lines. Tools like DNA Painter and Genome Mate Pro. If you haven't already, I'd surf around sites specializing in use of DNA information with adoptions, like DNAadoption.org. In general, working to identify biological family with autosomal DNA is less complex than trying deep dives back to, say, 5g-grandparents.

Mitochondrial DNA--which isn't part of the human genome at all--was the first type to be fully sequenced, and really the only kind up until relatively recently (as in the last decade) used to test ancient remains. The simple reason is that the tiny mitochondrial DNA molecule (only 16,569 base pairs all told, compared to the smallest human chromosome, number 21 at 48 million base pairs) was easy to sequence, and because each human cell (except red blood cells) have scores or even several hundred mitochondria in them...you have over a quadrillion mitochondria in your body. That makes for a lot of copies to test if some are damaged; ergo the usefulness in looking at ancient remains before HiSeq testing started taking over that realm circa 2014. Now mtDNA testing of ancient remains is a backup plan only, seldom the first type of sequencing attempted.

Point being, mtDNA is extremely valuable if trying to disprove a specific hypothesis, but not so useful genealogically at proving a hypothesis or randomly finding matches to unknown cousins. Because the mtDNA molecule is so tiny--yet has half as many coding genes as the Y-chromosome that comes in at about 58 million base pairs--there simply isn't much room for mitochondrial DNA to mutate. Even with an exact match on a full-sequence test, FTDNA gives only a 50/50 probability that you share an ancestor within 5 generations...and a lot of very bright people, like Dr. Ann Turner, believe that probability is way too lenient based upon numerous scientific studies of mtDNA mutation rates.

That's why mtDNA for genealogy can be misleading. Living DNA and others tout that they're identifying your maternal line, but in fact they're only testing select SNPs in order to arrive at a solid estimate of your haplogroup. That's great information, and it can tell you definitively that you don't share an ancestor with someone else; but it can't provide any meaningful evidence that you do share an ancestor in the genealogical timeframe.

The math there is pretty straightforward. The keeper of the mtDNA phylogenetic tree, phylotree.org, where all the haplogroups are cataloged, has just over 5,400 "nodes" defined in their most recent Build 17. A "node" is a defined ancestral haplogroup. For example, H4a1a1 is a node, and so is H4a1a2.

If you assume 5,500 mtDNA nodes, and a global population of 7.6 billion, you'll see that there are almost 1.4 million people for each mtDNA haplogroup. Really common ones like H2 will have many times that number, and very rare haplogroups will have markedly fewer. But still, the population of San Diego, California, is 1.39 million. So knowing your mtDNA haplogroup--even if it's refined down to the level of H4a1a1--would still be the equivalent of saying you've established DNA matching evidence to someone simply because you both live in San Diego. wink

Heck, as of three weeks ago mitomap.org had 46,092 mtDNA full-sequences on file, likely second only to FTNDA. Even if each was a unique haplotype, each would have about 165,000 exact matches (averaging) in the world. And if you want to add a real kicker to that, the modern human mitochondrial genome and the Neanderthal mtDNA genome differ on a total of only about 200 base pairs. That's how slowly it mutates and how little variation there is among humans.

Sorry for the digression. Something of a soapbox issue. Bottom line, though, is that autosomal DNA is always the best first testing choice if you're just seeing what you can see or if you're looking for close relatives. Both yDNA and mtDNA--unless you just want to spend the money out of curiosity--really require going into the testing with a plan and with hypotheses in hand. The yDNA and mtDNA haplogroups that 23andMe and Living DNA provide are only afterthoughts because the microarray testing chip they use, the GSA chip from Illumina, tests some yDNA and mtDNA SNPs by default and can be programmed (customized) to test for a few more. It's one testing pass on one microarray chip; they don't do for-purpose testing like FTDNA or YSEQ, for example. The 23andMe and Living DNA tests can provide a haplogroup estimate that's very accurate, but doesn't tell you much genealogically speaking.

As you can no doubt tell, though, it's all cool stuff to me, so I test everything I can test. Except full-genome sequencing. Not quite ready to shell out even $1,200 for that yet. I will once it's under $1,000 and becomes more popular..and somebody figures out a way to manipulate the massive raw data files for matching and comparison. smiley

"One DNA expert told me that 3 siblings will get a good picture of the grandparents."

Ooh. Doug reminds me. I don't often run into folks who have themselves plus at least two full-siblings to work with, but on the off-chance you do, Nancy...

There's a technique called visual phasing that can be used to--at least partially--reconstruct what the DNA of all four grandparents looked like. There's a Facebook working Group for it: https://www.facebook.com/groups/visualphasing/. If you join that Group, you'll find a dozen files in the private area with everything from instructions, how-to info, and an Excel template to use to get started. Takes some work, but you learn a lot in the process and with enough data to throw at it you can come away with a good mechanism to sort all your current and future DNA matches into each of the four grandparent's lines.

Thank you Doug and Edison. You both certainly seem to be fans of FTDNA, so I will look closer at that service.

Is there any way I can upload my existing DNA results to FTDNA? I came across one possible solution....that talked about tweaking the Ancestry file to make it uploadable to FTDNA. I'm not sure if it works with 23andme V5 as well. FTDNA forums seem to have been down for a fair period of time so I can't ask the specific person about it yet.
Yes, you can upload DNA for free but some of the features require paying a one time fee of $19. Go to FamilyTreeDNA.com and click on Shop. Then go to the end of that page.

Nancy, right now getting 23andMe v5 results uploaded to FTDNA is a no-go. Same for Living DNA results. The issue is that 23andMe and Living DNA are using a newer microarray chip from Illumina called the Global Screening Array (GSA) chip. FTDNA, AncestryDNA and, for now, MyHeritage all use the OmniExpress chip. One isn't inherently better than the other, but Illumina saw the GSA chip as the best move forward, commercially.

The GSA chip can test a total of 665,608 SNPs; the OmniExpress 713,599. Neither is inherently better than the other for genealogy. The twist--and why you can't upload your 23andMe raw data to FTDNA, and the reason it's taken far longer than expected for GEDmatch and GEDmatch Genesis to merge databases--is which SNPs are tested. You see, the two tests share only about 23% of the same markers. So over 3/4 of the SNPs examined by each chip type are unique to that chip.

Makes it very difficult to accurately compare results from one test with the results from another...and whole lot of fancy math and imputation is involved to make it happen. Frankly, I believe the verdict is still out as to exactly how accurately this can be accomplished right now. With MyHeritage already accepting both tests into a single database, I'm sure we'll see reports about experienced accuracy in the upcoming weeks.

I'm uncertain what the issue would have been with the AncestryDNA-to-FTDNA file conversion you mentioned. AncestryDNA raw data have been uploadable to FTDNA since v1, and they still use the OmniExpress chip now, so current Ancestry test info is still good to go.

It is possible to edit the raw data file if you know what you're doing. It's just a plain-text ASCII file. But the specific layout and spacing is critical. Each tested SNP is known as a reference sequence or reference cluster. If you open your own data file you'll see those 650,000+ lines each preceded with the letters "rs" and a number. All those numbers are unique and you could theoretically remove the 77% that are unique to the GSA chip and be left with reference clusters that FTDNA's database understands. The catch here is that you'd end up with a file containing information for only about 155,000 SNPs, and FTDNA would determine it to be an invalid kit because it had so few markers tested.

Just a FWIW, on the flip-side of that situation is the fact that GEDmatch Genesis is very liberal in the data files it can accept. Some enterprising genetic genealogists have merged test info from an OmniExpress chip and a GSA chip to create results data that contains over 1.1 million SNPs, with the--probably correct--notion that more is better when it comes to the comparison engine in Genesis. To my knowledge there is, as yet, no online utility that will allow you to upload one of each type of test and give you back a properly merged file. But ya just know somebody's going to do it.  cool

Thanks for this incredible information Edison. I'm not sure I'm ready to absorb it yet as I want to crawl before I walk, but it reminded me how much I loved my genetics class in university so many years ago.

And, regarding your latest comments on FTDNA, that is really helpful. I think that's a dead end for me now, so nice to know that I shouldn't expend more energy there.

Now, if only EVERYONE would upload their DNA to GEDmatch Genesis!
You say that I can upload the file, however it didn't work when I tried it and the FTDNA website says 23andme v3 and v4 can be uploaded. I have v5.

Do you know of anyone who has successfully uploaded their DNA file from 23andme V5?

I trust FTDNA primarily because they did a good job for me, told me that my missing GF (on dad's side) was Jewish, and that they don't keep on charging you and keep on keeping-on in that matter till perpetuity. They really do help with pointing you toward making a gedcom. In Addition: they only do testing and matching. They do have a tree but it's casual, i.e., a convenience for us because many of us don't have a tree for a first time DNA testing situation. If you want the best place for a historic-records-based tree after you've been tested, it's Wikitree, an international collaborative site for everyone, and it is free of charge. If you want records, then Familysearch.com is the go-to source (also free of charge). Wikitree members want you to produce clean and correct trees and they'll direct you to write your sources as you go, but as many of us with years of trying still have trees with errors and almost errors. That's where the collaboration really helps.

Have a happy, enjoyable and sometimes difficult adventure with your genealogy. I feel more whole every year that I engage in it.

+3 votes

My own experience has led to a lot of mixed feeling about your dilemma.  Ancestry has the most other testers to match with, but they provide very few details (no chromosome browser).  They can be considered the universal donor.  Their results can be transferred everywhere else except 23andMe, but they accept no transfers in.  For this reason alone, if you can only do one test, you should consider them.  It really bugs me to say that because of their attitude of "we're the big dog, so we don't need to cooperate with others or improve our service."

23andMe is kind of a stand alone service, but they have the second highest number of testers.  I've found a high percentage of my matches there that aren't found on other services.  Their user base isn't very cooperative, so sometimes it takes a LOT of detective work to figure out the other tester's ancestry.  As you found, until recently, their test could only be uploaded to GEDmatch Genesis.  The GEDmatch database merger has been coming soon for quite awhile.  Just recently, MyHeritage announced that they will accept v5 tests from 23andMe and Living DNA.  See MyHeritage supports 23andMe V5 and Living DNA uploads for recent discussion.  Note that this remains completely free only until December.

LivingDNA  also now accepts transfers from any other testing company.  See this link for details:  https://www.livingdna.com/help-centre/332/uploading-data.  They're still in beta testing and haven't produced any matches yet.  I would expect that they will remain the smallest player for the foreseeable future.

I should say that my favorite site for DNA matching is MyHeritage.  They automatically show triangulated matches.  Many of their testers have posted a family tree, so it's often possible to find the MRCA.  They can now be considered the universal recipient, which unfortunately makes testing on another site more logical.

FTDNA has not kept up with the times.  Their interface seems dated.  They provide no triangulation option.  Their only advantage, as Edison described, is the capacity to add other types of DNA tests at a later date.  I recently added a Y-DNA test from my father's sample which led to zero useful information.  The 23andMe paternal haplogroup was more useful to me, but, of course, this will vary from person to person.

My opinions.  YMMV.

by Kerry Larson G2G6 Mach 9 (99.6k points)
Thank you Kerry. Your comments about each of the services is very helpful.

I also thought of Ancestry as the big bully on the street. Of course, their approach is that 'you have to come here' if you want access to our tree database for your DNA matches.

I did upload to My Heritage and I did find a 2nd cousin who I suspect is on my mother's side as the related tree didn't have any names that I recognized at the ggp or 2ggp level from my father's side. I'm not sure how to take this any further though.

I'll likely need further matches to identify which is the specific common ancestor.

Having your father tested, on any site, will be the most useful thing you can do for finding DNA matches.  After all, he has twice as much DNA as you for his side of the family.  It'll also help you for your maternal side since whoever you're related to that he is not will obviously be through your mother.  I did have my own DNA tested but it was probably a waste of money.  I never look at my own results as I had both of my parents tested and only look at their results as I can't have any matches that they don't.  Good luck, whatever you decide. 

Thanks for the tips. I think you make a lot of sense and confirms to me that it's worth the money to get my father tested. I'm probably overthinking things, as I was disappointed that the 23andme results weren't as 'shareable' as I had thought they would be.
I think you'll be happy with the additional matches once you have your father tested.  You'll also eventually feel better about your 23andMe test.  There's a large chunk of testers there that you just won't find anywhere else, partly because they can't transfer, but more likely because they just did it for the ethnicity and/or health results.

To add to Kerry's excellent advice (and because everybody knows I'm never at a lack for words <cough>), once you get your father's data back there's a super-easy-to-use feature at GEDmatch that can help you immediately with locating matches on your mother's side of the family.

GEDmatch simply calls it "Phasing" (though it isn't traditional trio phasing). If you have the data on one parent and one child, you're good to go. Remember my yammering about the chipset differences? To get the best results with GEDmatch's phasing tool, you'll want both tests taken at the same company and with the same version of the test. Not 100% required, but that assures you start with the maximum number of the same tested SNPs.

The phasing tool is as easy as entering both GEDmatch kit numbers. In essence, what it does is look at each SNP, or reference cluster, in each kit. If a value at a specific position in your data cannot have come from your father then, voila!, we know it had to have come from your mother. The end result is that GEDmatch creates a new "pseudo-kit" that approximately represents what your mother's test data would look like. No one else can see that pseudo-kit in one-to-many searches, but you can use that way, and in one-to-one match examinations, as well. Pretty useful!

I say it approximately represents your mother's test data because there will be spots where the algorithm can't accurately make a prediction about the correct value for your mother's kit. One example is that every set of raw results has no-calls--loci where the test couldn't accurately determine a nucleic acid letter and the value in the file is a zero, a null value; this is almost always less than 3% of the markers in any kit.

The new pseudo-kit usually won't match cousins as closely as would have your mother's actual data because the new kit has fewer SNPs in it and can thus predict fewer shared segments. But from experience, this won't be farther off than a half relationship. In other words, if your mother has a 2nd cousin who has tested, this pseudo-kit may make the range of centiMorgan sharing look more like it's a 2nd cousin 1x removed, but it shouldn't be off farther than that. The relationship estimations (time to most recent common ancestor) will be off, but workable.

The really cool thing, though, is that any kit on GEDmatch that matches your mother's pseudo-kit, will not match your father's data. So can immediately tell on which side of the family to place that relationship. It's an important tool for adoptees who can test the known biological parent.

Have fun with the search!

I think there is great insight here Edison, but please bear with this newbie. When you refer to this pseudo-kit, what exactly do you mean? Are you implying that I would have to look at the specific chromosome mis-match between me and my father and then compare this mismatch with other 'GEDMatch relatives'? Is there a specific report that would highlight this or is it detailed chromosome analysis?

And, does this solve my problem with only having my DNA in GEDMatch Genesis? i.e. would it require me to take another DNA test that is compatible with the older/broader GEDMatch database?

Thanks for you patience in explaining this to me.

Sorry for the delay in responding, Nancy. Nope; the one-parent phasing won't solve the GEDmatch Genesis vs. GEDmatch original problem for you...at least not until they can effectively merge the two databases. And I should have been more clear: the phasing tool at the standard GEDmatch is free, but at the GEDmatch Genesis site it's part of the Tier 1 utilities. You can get access to Tier 1 tools for $10 per month, and you can stop the subscription at any time; so you can just get it for one month if all you want to do is perform some specific functions, or let the subscription continue as long as you want.

As to the pseudo-kit produced by the computational phasing, it's a real kit with its own kit number. In fact, two pseudo-kits are always produced, but since you will have your father's real data you won't need that one. If your 23andMe data on GEDmatch Genesis is kit number M12345 (example only), the two kits generated will be M12345M1 for your mother, and M12345P1 for your father.

After you run the phasing tool, it will take about a day for the pseudo-kit to "tokenize." After that, you can use it just like any other kit number. Just plug M12345M1 into whatever GEDmatch tool you're using. You can generate a one-to-many report from it, or use it for a one-to-one comparison with a found match and generate the same kind of detailed chromosome analysis you see when using your own kit number. That's why it's so valuable for adoptees: it's simple to create; it stays on file and is reusable for any purpose; and any matches you find with it will be exclusively to your mother. The only difference between it and an uploaded set of data is that other people won't find the pseudo-kit when they run one-to-many type reports. It's hidden from searches; you have to know the kit number in order to use it.

One thing I didn't mention, and it's always a first-step I recommend when a new kit is upoladed to GEDmatch. It puts some folks off when I mention it, but it's an important place to start: run the "Are Your Parents Related?" report. Run it on your own kit number now, and on your father's once his data is uploaded.

All this tool does is check a kit's raw data for what are called runs of homozygosity (ROH): blocks of a chromosome where the nucleic acids are the same. If you look at your raw data, for any given reference sequence you might see an A and an A, or a C and a C. This means that you positively know that, in the first instance, both your mother and your father gave you the nucleotide Adenine at that locus; both gave you Cytosine in the second example.

If there are relatively large, unbroken stretches where both parents contributed the same "letter," it's a solid indication of pedigree collapse, of your parents sharing one or more a biological ancestors. This may have some impact on the amount of DNA you might expect to share with certain cousins but, more importantly to adoptees, if the shared ancestors are relatively recent it may affect the ability to neatly sort matches into the paternal and maternal lines. The key word is may.

But the "Are Your Parents Related?" tool at GEDmatch will give you a good indication if there is any additional relatedness to be aware of. GEDmatch's tool is based on the work of a mathematician in Newfoundland, David Pike. You can't adjust the parameters of the tool at GEDmatch, so if you do find there is pedigree collapse to be concerned with, you might want to investigate the utilities that David provides in order to dig a little deeper: https://www.math.mun.ca/~dapike/FF23utils/.

Most will never have to worry about it, but it's better to know up-front and to plan accordingly than to spend hours and hours trying to figure out why you're finding matches that seem like they might be on both your parents' maternal and paternal lines, or the same situation on a grandparents' lines. wink

Related questions

+4 votes
2 answers
+17 votes
3 answers
+9 votes
1 answer
+2 votes
1 answer
253 views asked Dec 28, 2018 in Genealogy Help by Billy Dunn G2G6 (7.3k points)
+5 votes
1 answer
+6 votes
2 answers
612 views asked May 13, 2018 in The Tree House by Azure Robinson G2G6 Mach 9 (94.9k points)
+19 votes
1 answer

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

disclaimer - terms - copyright

...