PDA

View Full Version : MtDNA matches, more particularly



Judith
09-18-2016, 09:44 PM
Ftdna say that if you match exactly on their full mtDNA test then there is a 50% chance that the common ancestor is within 5 generations, 125 years. Roberta Estes calls this value a lot optimistic in her blog.

I have been doing a lot of reading of the literature about mtDNA mutation rates, the differences between pedigree derived rates and phylogenetic rates. I can not find a single paper where the median time for one mutation is 125 years or 5 generations. I can find 50-150 generations for one mutation using pedigree measurements.
What am I missing here?

Judith
09-19-2016, 07:24 AM
The paper that I have used most is Howell, (see below and followed up the references and citations). Table 2 gives a listing of publications to this date (2003) and the median is 100 generations with the +/_ one std deviation as 50 to 150 generations for one mutation.

The Pedigree Rate of Sequence Divergence in the Human Mitochondrial Genome: There Is a Difference Between Phylogenetic and Pedigree Rates

Neil Howell1, 2, 3, , , Christy Bogolin Smejkal3, D.A. Mackey4, P.F. Chinnery5, D.M. Turnbull5, Corinna Herrnstadt1
Show more
http://dx.doi.org/10.1086/368264

Is this work generally accepted?

Jenny
09-19-2016, 07:57 AM
My mother is from Finland, she has five exact full sequence matches, two from Finland the rest Sweden and Norway.

Finns are avid genealogists and while we match geographically, we cannot find a historical woman in common even though our trees go back to the 16 and 1500s.

Hope this helps!

Judith
09-19-2016, 11:45 AM
Jenny, you are far from alone!
Many people are now finding that their full sequence and exact matches does not lead them to a common woman in 20 generations like your example. I was trying to find out why, starting with where ftdna got their prediction from. I am struggling to find anything remotely like 5 generations in the scientific literature for one mutation. The lowest figure published is ~30!
My apologies everyone the median is 100 GENERATIONS it was a Typo above, and I have now corrected it in the earlier post.

Lirio100
09-19-2016, 02:19 PM
Same thing; I have four exact matches and can find no connection other than geographical.

Wing Genealogist
09-19-2016, 02:50 PM
FTDNA's algorithm for mtDNA matches assumes you do not know whether the two individuals are siblings (with the same mother). I believe it does have the ability to add in the number of generations you know your match is not related to you.

Even given the above, FTDNA's matching for mtDNA does need to be fixed.

C J Wyatt III
09-19-2016, 03:00 PM
Same thing; I have four exact matches and can find no connection other than geographical.

I suspect that there is a lot faulty with the knowledge of mtDNA. My belief is that mutations and back-mutations occur at a far higher rate than is generally thought. When I am able to find autosomal kits to go with mtDNA results, I see a lot which points to a connection within the genealogical time frame.

Here is something you can try at home (if you have your kit on GEDmatch). Take your kit (or better one belonging to parents, grandparents, etc.) and run an 'X one-to-many' with default parameters. Next, on the mtDNA column click the down arrowhead to sort by haplogroup. Go through the resulting list and look at cases of where you have an X-match with two or more from the same mtDNA subclade. You will get a bunch of instances like that. The only answer that I can think about those little groupings is that you are triangulated with those kits somewhere on the X-chromosome (and given the size of the match, it is more recent than distant).

Jack Wyatt

Ann Turner
09-19-2016, 05:27 PM
Lately I've been using the mutation rate 6 * 10^-7 (per base per generation) from Kivisild "Maternal ancestry and population history from whole mitochondrial genomes."

http://investigativegenetics.biomedcentral.com/articles/10.1186/s13323-015-0022-2

I agree that FTDNA is far too optimistic about the TMRCA.

Judith
09-19-2016, 07:27 PM
I have had a browse with the X chromosome but the snag is you have 100 to 1000 possible women for the X. I have a match at 7cM with one lady and it may be 6 generations ago or it may be 10. Right now it is filed mentally under too difficult, since few parts of my tree go back 10 generations, and because of the nature of parish records it is the women who are the least known.

GailT
09-20-2016, 05:07 AM
The FTDNA FAQ is not accurate and needs to be updated. It has not been revised in over 6 years, and I don't think it was based on any published research. I copied below my standard response to questions about mtDNA matches.

The mtDNA mutation rate is slow, an average of about 1 mutation in 3600 years (using the Soares estimate), but it is also highly variable. For some people with no "recent" mutations, their exact matches could share a common maternal ancestor several thousand years ago. For people with recent mutations, exact or even close matches can be much more recently related.

I posted a while ago about a person who is a 3 step match with her mother and a 4 step match with her sister (all 3 tested the FMS). This distance among close relative is extremely unusual, but it highlights the fact that you should view your match list as possible lines of research, especially those matches (at any distance) who are from the same areas as your maternal line.

If you have a large number of FMS matches, it is likely your line does not have recent mutations and most of those matches will be very distantly related. If you have a small number of FMS matches, you should look at them more carefully for possible connections.

Also, HVR matches can be very useful in cases with a recent mutation in the HVR region. For people who have a small number of HVR matches, you should check for any matches who have maternal ancestry from the same area.

A mtDNA match is not proof of a recent maternal line relationship but it can confirm or disprove suspected relationships. It can also identify a group of people with a small number of matches who are likely to have a common maternal ancestor within the last several hundred years, and that group could then work together on research on their maternal ancestry. As more test results become available, we can get much more value from mtDNA if more people would share info on their maternal ancestry and if they understood how to interpret and use the results.

Judith
09-20-2016, 04:58 PM
Thanks Gail, that is better written than what I would have put! I concluded that the key problem was the marketing people wrote the FAQs and 5 generations is about where many people find their trees get fuzzy. I expected a factor of two to 5 between the marketing and the science however, not a factor of 20. Everyone's experience seems to be the same and at a guess it seems that one person in ~100 or less finds their mrca with a 100% full match.
Whether this is back mutations on the hot spots or other sites is an open question.
It does not seem to be a question of correcting an algorithm but of managing expectations with the site information.

Huntergatherer1066
09-20-2016, 10:40 PM
My advisor in grad school Lorena Madrigal was keen on studying mtDNA mutation rates; she did an interesting study in a Costa Rican village comparing mutation rates with pedigrees from the local church. I had the honor of designing the poster for the paper and transporting it to Portland for the 2012 AAPA meeting in Portland. Here is her paper:

Madrigal L., Loredana Castŕ (Posthumously) , Mauricio Melendez-Obando, Ramon Villegas-Palma, Ramiro Barrantes, Henrieta Raventos, Reynaldo Pereira, Donata Luiselli, Davide Pettener, Guido Barbujani. 2012. High mitochondrial mutation rates estimated from deep-rooting Costa Rican pedigrees. American Journal of Physical Anthropology. 148 (3): 327–333. DOI: 10.1002/ajpa.22052 http://onlinelibrary.wiley.com/doi/10.1002/ajpa.22052/abstract

Saetro
09-21-2016, 12:51 AM
Lately I've been using the mutation rate 6 * 10^-7 (per base per generation) from Kivisild "Maternal ancestry and population history from whole mitochondrial genomes."

http://investigativegenetics.biomedcentral.com/articles/10.1186/s13323-015-0022-2

I agree that FTDNA is far too optimistic about the TMRCA.

Thank you so much, Ann.
This paper is exactly what I have been looking for.
In particular it highlights the differences between mutation rates for different places on the genome, whereas FTDNA is using some sort of overall figure - and that probably refers to an old number for HVRI and or HVRII.
I can now go back to matches and work to estimate better when our TMRCA might be.

Caveat 1
These are averages. Statistical averages apply to populations. It is the spread of results that is more relevant to individuals.

Caveat 2
We need to know both our own result and that of our match(es) to apply these mutation rates.
As even FTDNA mtDNA projects do not show the coding region, this requires close cooperation - and understanding - from matches.
Until recently I had found matches (even newbies) understood that mtDNA was only transmitted along the all-female line.
Now (and perhaps this is just me) I find recent matches referring me to surname studies, or giving their oldest known mtDNA relative as someone with the same surname as their maternal grandmother. (Back in the 1200's!!!)
I need to revise my expectations of how to work with mtDNA matches and update my recommended references.

FTDNA should certainly update its advice - if only to refer to the ISOGG wiki on mutation rates http://isogg.org/wiki/Mutation_rates which should be updated to include this Kivisild paper among the references.

geebee
09-21-2016, 02:04 AM
I have seven FMS matches at FTDNA. Two have a genetic distance of three; two have a genetic distance of two; one has a genetic distance of one; and two have a genetic distance of zero.

Only the two with a genetic distance of zero appear to be related to me. I'll use their initials, WER and JSK. JSK actually lists the same woman I do as his "most distant [mtDNA] ancestor": Marie Catherine Anne Berda dit Picard. WER lists a different woman, Angelique Ladner. But Angelique Ladner was Marie Catherine Anne's daughter, so WER could go one more generation back.

Both WER and JSK either took the Family Finder test or uploaded a file from another company. However, only JSK appears as one of my FF relatives. WER does not, so it looks as if Angelique might be our most recent common ancestor. I'm not sure what generation back she is for WER, but for me it's seven generations.

Since our GD is 0, it would appear that neither WER nor I have had any mutations in at least that number of generations.

JSK is quite a bit more closely related to me. His maternal grandmother and mine were half sisters, so JSK and I are half 2nd cousins.

Judith
09-21-2016, 08:04 AM
Thanks HG1066 this is one of the now many papers which indicate ~100 generations

Ann Turner
09-21-2016, 02:12 PM
I suspect that there is a lot faulty with the knowledge of mtDNA. My belief is that mutations and back-mutations occur at a far higher rate than is generally thought. When I am able to find autosomal kits to go with mtDNA results, I see a lot which points to a connection within the genealogical time frame.

Here is something you can try at home (if you have your kit on GEDmatch). Take your kit (or better one belonging to parents, grandparents, etc.) and run an 'X one-to-many' with default parameters. Next, on the mtDNA column click the down arrowhead to sort by haplogroup. Go through the resulting list and look at cases of where you have an X-match with two or more from the same mtDNA subclade. You will get a bunch of instances like that. The only answer that I can think about those little groupings is that you are triangulated with those kits somewhere on the X-chromosome (and given the size of the match, it is more recent than distant).

Jack Wyatt
I do agree that it would be great to have more mutation rate data based on deep-rooting pedigrees, like the Madrigal study Huntergatherer1066 posted. It's just so complicated to study, because of hteroplasmy and the extreme rate heterogeneity for different positions etc.

I did a very quick check of my son's X one-to-many matches at 7 cM, but it would take a lot of effort to look at the triangulation aspect. I do have a couple of quick observations. He had 78 and there were indeed little clusters of people with the same haplogroup. But a lot of them had the same email address, so they were probably close relatives. That wouldn't count as triangulation.

Another fact jumps out as a dramatic demonstration of phased vs unphased matches. My son's X is automatically phased since he has just one X. My X match lists maxes out at 2000 by the time it gets to 7.6 cM. My son should about half of those if they were true matches, so my list must include a lot of false positive pseudo-segments.