Effect of additional markers on most recent common ancestor estimate

[J Honeychuck]

People have asked how the estimate of Time to Most Recent Common Ancestor changes when additional markers are purchased for the pair of cases. Here is a case from my research. I don't assert that this is typical, just a real case.

The Y-DNA haplogroup is J1* M267. Both are European cases. All estimates using http://dna-project.clan-donald-usa.org/tmrca.htm

Number of matching markers and number of markers tested
Most likely number of generations back to the common paternal ancestor
Estimate by time using 25 years per generation

10/12 -- 45 generations -- 1,125 years, 875 AD
21/25 -- 38 generations -- 950 years, 1050 AD
30/37 -- 32 generations -- 800 years, 1200 AD
41/50 -- 30 generations -- 750 years, 1250 AD
57/67 -- 29 generations -- 725 years, 1275 AD

[nathanm]

People have asked how the estimate of Time to Most Recent Common Ancestor changes when additional markers are purchased for the pair of cases. Here is a case from my research. I don't assert that this is typical, just a real case.

The Y-DNA haplogroup is J1* M267. Both are European cases. All estimates using http://dna-project.clan-donald-usa.org/tmrca.htm

Number of matching markers and number of markers tested
Most likely number of generations back to the common paternal ancestor
Estimate by time using 25 years per generation

10/12 -- 45 generations -- 1,125 years, 875 AD
21/25 -- 38 generations -- 950 years, 1050 AD
30/37 -- 32 generations -- 800 years, 1200 AD
41/50 -- 30 generations -- 750 years, 1250 AD
57/67 -- 29 generations -- 725 years, 1275 AD

That's not how that tool was designed to be used. You must be clicking the "Create Graph" button, which looks like a normal distribution with the peak at the TMRCA, but isn't.

First, the tool shows the number of "transmission events," which are twice the number of generations. It's confusing, but that's how Dr McDonald designed it. Also, 25 years per generation is low, a better estimate is about 30 years. That takes into account the average age of the parents at the birth of all their children. Many studies have been conducted on just this question, with varying lengths. For one thing, the average generation length is higher for men than women, and varies significantly by culture, religion, time period, family, etc. Recent studies are resulting in even higher numbers, between 31 and 34.

Second, the default mutation rate of 0.0033 is the calculated, overall rate for 37 markers, although Dr McDonald expresses some skepticism for the rates above 12 markers in the text at the top. If you want to calculate for a different number of markers, you should change the mutation rate to the one listed. And better yet, if you know which markers the two samples differ on, look them up and see if they're significantly higher or lower than the overall average (my preferred mutation rate listing is this color coded table). If they were different on only one marker, it would be very easy to use that figure. However, differences on more than one marker complicate the situation. Averaging the mutation rates for those specific markers isn't technically correct, but it may give a better ballpark figure than using the overall average.

Third, if you look at the probability on the Y-axis of the graph for 10/12 markers, it's only 0.0098 at 45 transmission events. In other words: a 0.98% chance there were exactly 22.5 generations between the samples and a common ancestor. Not anything most likely, or even likely, about those odds. And it doesn't mean less than that number, or around that number, but only exactly that number. Instead, make sure to click the "Cumulative Probability" check box. Then the results show the probability of a common ancestor in that many transmission events or fewer. Then, report the number of generations (divide transmission events by two) where the probability breaks some easily imaginable figure, like 0.5 (50% chance) and 0.9 (90%). It's also a bit easier to read from the list, rather than the graph.

Now, using the default, overall mutation rates for each set of markers, the estimates for the MRCA from your example are quite different:
10/12: 50% chance within 61 generations, 90% chance within 122 generations
21/25: 50% within 44, 90% in 76
30/37: 50% within 35, 90% in 53
57/67: 50% within 31, 90% in 45

[1798]

Your post makes a lot of sense. I have 1100 11/12 marker matches from 40 Countries. Most of these people are not in the same subgroup that I'm in.My haplotype is 100% R1b. Our common ancestor must be M269 and about 8-10,000 ybp.I am interested in the man from Turkey that matches me. If he is aboriginal Turkish then he is descended from the R1b clan that stayed in Turkey 25,000 years ago and we still have 11 of the same values in the first 12 markers!!