FOX SURNAME Y-DNA PROJECT
Progress Report No. 3
December 6, 2004
Results for nine project members are now available from FTDNA. The main developments since Project Report 2 are that we now have two brothers tested at 25 markers and a father-son pair tested at 12 markers. Both pairs showed identical results.
Results have also been obtained for a British Fox family member who was tested by Mark Jobling at Leicester University. He shows a 10 for 12 correspondence with the first two project members but the DYS Numbers tested are not the same as for the FTDNA 12-marker participants. His ancestry is said to date back to a brother of Sir Stephen Fox. He has now joined the project at FTDNA and it will be interesting to see whether FTDNA obtains the same results and whether a null value is obtained for DYS439.
Terry Latey and Gary Patrick Foxx have given permission to release their test results and ancestry information to the Fox mailing list in the hope that someone will see a connection and join the project. There was only a one-step mutation on one marker between Terryâ€™s father, Thurman Eugene Fox, and Gary in the 12 marker test. Gary traces his ancestry back 8 generations to Wuerttemburg, Germany, while Terry is stuck at 5 generations in Adair County, Kentucky, but feels that her fatherâ€™s ancestry is also German.
The other news is that FTDNA has just released average marker mutation probabilities for their 12, 25 and 37 markers tests, which are based on an in-house analysis of family relationships (in databases such as Ysearch) done by a University of Arizona student. These probabilities are typically 1.5 times higher than those they were using before, thus reducing the estimated number of generations to a common ancestor. They have provided Group Administrators with a proprietary patented tool, called FTDNATiPÂ®, which predicts the probable time to a common ancestor for any pair in our Fox Surname Project. They say that this will soon be made available to all participants. FTDNAâ€™s predictions for the two close matches in our Fox Project have been graphed and are reported below. In addition, the Poisson Model analysis given previously in Technical Report 1 has been extended to show the effect of these new mutation rates. Until these new rates are officially published and given proper peer review, however, they should be considered very tentative.
Table 3.1 is a summary of the test results to date, with the participants identified by kit number and most distant known ancestor. The markers (identified by DYS Number) are specific locations (microsatellites) along the Y-chromosome where the alleles repeat themselves and the numbers shown are the number of repeats at each marker for that participant. The sequence of numbers for each individual is known as his haplotype. The results have been grouped by known ancestry and haplogroup. For haplogroups R1b and I, the modal average haplotype has also been shown and deviations from this average are highlighted. The derivation of the modal values can be found at: http://worldfamilies.net/Super%20Western%20Atlantic%20Modal%...
The other change that has been added is the classification of markers as slow, moderate or fast mutators by color-coding. Individual marker probabilities were discussed in Technical Report 1. While individual marker rates give a feel for whether a result is reasonable, what really matters in assessing probabilities of a relationship is the average probability of mutation for all the markers tested. FTDNA released the following values at a recent meeting of their participants: First 12 markers, p = 0.00399; next 13 markers, p = 0.00481; last 12 markers, p = 0.00748. From these values, I have calculated the average mutation probability for the 25 marker test to be 0.00442, while for the 37 marker test it is 0.00541. (In contrast, FTDNA reportedly gave the average for 37 markers as 0.0058.) These probabilities are much higher than the averages of rates given earlier at http://worldfamilies.net/marker.htm
. and used in Technical Report 1.
It is obvious that FTDNA has deliberately picked slow markers for their 12 marker screening test and markers that are more likely to mutate for their 25 and 37 marker tests. Note that a probability of 0.004 corresponds to one mutation every 250 events, whereas a probability of 0.0075 corresponds to one mutation every 133 events. A father and son would have one event, whereas brothers would have two events separating them.
In Technical Report 1, the Poisson Distribution for Rare Events was used to predict the probabilitiy of 0, 1, 2, 3, 4, etc. mutations in the 20 transmission events (10 generations in each line) expected for the first two participants, 14179 and 16564. Recall that there were 2 mutations in the 25 marker test and four mutations in the 37 marker test. The old mutation rates were used.
Results were as follows:
25 marker result: 0 mutations = 24%, 1 = 34%, 2 = 24%, 3+ = 19%
Two or more mutations are expected 42% of the time.
37 marker result: 0 mutations = 12%, 1 = 26%, 2 = 27%, 3 = 19%, 4 = 10%, 5+ = 6%
Four or more mutations are expected 16% of the time
Using FTDNAâ€™s new mutation probabilities, these results change considerably:
25 marker result: 0 mutations = 11%, 1 = 24%, 2 = 27%, 3 = 20%, 4 = 11%, 5 = 5%, 6 = 2%
Two or more mutations are expected 65% of the time and 2 mutations is the most likely result.
37 marker result: 0 mutations = 2%, 1 = 7%, 2 = 15%, 3 = 20%, 4 = 20%, 5 = 16%, 6 = 10%, 7 = 6%
Four or more mutations are expected 57% of the time. Three or four mutations are the most likely results.
Probabilities for the most recent common ancestor (MRCA) were also predicted using FTDNAâ€™s new tool, FTDNATiPÂ®. The results are plotted in Figure 3.1 for the same two participants. Curves for â€œ25 markersâ€ with two mutations and â€œ37 markersâ€ with 4 mutations are shown. For both curves, the cumulative probability for a common ancestor 10 generations back is found to be 40%. Adding more markers to the test made the ultimate probability of a relationship more convincing.
FTDNATiPÂ® also allows the known prior information to be input that the two individuals did not have a common ancestor for 9 generations. This additional information is reflected in the lower curve shown in Figure 3.1, labeled 37 markers (without the quotation marks.) The tool now starts from scratch at 9 generations and the probabilities are renormalized to end up totaling 100%. The curve shows about a 10% probability of a tenth generation common ancestor. (This needs to put into context, however. Using the same procedure with a perfect 37 for 37 match would give about a 33% probability of a 10th generation common ancestor.) The two individuals are certainly related if one goes back 20 generations but that is about all one can say with certainty.
The effect of a single mutation in a 12 marker evaluation is shown in Figure 3.2. With no mutations (14179 vs 16564), a MRCA at 14 generations back has a cumulative probability of 75%. (The same would hold for 14179 vs 25549, or 24049 vs 24972.) At 25 years per generation, this would be 350 years. With one mutation, (24105 vs 24157) the same cumulative probability is achieved at about 27 generations (or 675 years.) A possible connection is remote but is certainly worth following up.
These MRCA predictions are useful tools for understanding relationships but are not precise by any means. For one thing, individual marker mutation probabilities are still being developed and FTDNAâ€™s rates are higher than others have published. For another, the method uses a Bayesian probability distribution, which is dependent on prior knowledge of probabilities of relationships within a large homogeneous population. Finally, if the calculation is based on â€œinfinite allelesâ€ it does not take into account the possibility of two mutations negating each other. Y-DNA testing is proving to be a valuable tool for denying a relationship but proving a relationship will always be a matter of assessing probabilities. The more participants, particularly known distant cousins, the better the estimate of the MRCA can be. Hopefully, a point of divergence from a common haplotype can be isolated if enough family members are tested.
Table 1 also shows the genetic heritage of all participants by classifications known as haplogroups. Haplogroups shown in red are considered firm estimates by FTDNA, values in black are less certain and could be checked out by further testing (a procedure known as SNP testing) if this is considered important. This was discussed in Progress Report 2.
The next month should see 25 marker results for a nephew, a grandson and a possible distant cousin of present project members. Participant 24157 has upgraded from 12 to 25 markers and his results should be available since they are now overdue. The father-son pair (24049 and 24972) has just ordered an upgrade from12 to 37 markers. Three project members have received test kits but have yet to return them. One of these is from Virginia, another is from Massachusetts and another is from the United Kingdom. The project seems to be proceeding well but more participants would certainly be welcomed.
Fox Surname Project Administrator