Is Your Family Tree Biologically Correct?

You know you’ve wondered; we all have:  Is my family tree biologically correct?  All of it?

I can easily prove that my social parents are also my biological parents because both have done DNA tests and match me as expected.

I can just as easily conclude that all four of my grandparents are who I think they are thanks to DNA matches to an uncle and some closer cousins.

But my confidence level decreases with each generation back as the DNA matches become more distant and the shared DNA amounts less predictive of specific relationships.  For example, I have a 2nd cousin once removed (2C1R) who shares only 19 cM with me.  According to the Shared cM Tool at DNA Painter, it’s far more likely that they are a distant cousin than 2C1R.  Is my tree wrong?  Is theirs?  Or is our match just an outlier.

 

To answer these questions, I need more match data for more people.  I also need a way to analyze all of that data at the same time.

Tree Validation with BanyanDNA

This is where BanyanDNA shines.  BanyanDNA is a new tool for genetic genealogy that is customized to your family tree, including cases of pedigree collapse and double cousins.  It has three main uses:  visualizing complex trees, validating biological relationships, and hypothesis testing for unknown parentage.  This post will showcase the validation features.

(Full disclosure:  I am a partner in the BanyanDNA business.)

BanyanDNA is unique among all genetic genealogy tools in that it can analyze multiple DNA kits at the same time.  This is fabulous news for those of us who manage the DNA results of other relatives.  BanyanDNA can even be used for match data from sites like MyHeritage and 23andMe, which show us how much DNA our matches share with one another, even when we don’t have direct access to their full match lists.   (Heads up:  AncestryDNA recently announced that they will soon offer us this information, as well!)

For example, on one side of my family, I have access to my own kits at AncestryDNA, 23andMe, and MyHeritage; my parent’s at 23andMe and MyHeritage; and a 1C1R’s at AncestryDNA.  I can plug all of that match data into a single analysis to see how it all fits the tree.

The images below show the tree in BanyanDNA from the perspective of each of the three of us.  In the upper image, you can see that I have match data (shown in the small “flags” at the top left of each of match) to everyone who has tested (the purple nodes).  The second panel shows a single match for my parent, who is only at 23andMe.  The third panel shows my 1C1R’s matches at AncestryDNA.

Once I have the tree and enter the data, I can ask BanyanDNA to analyze all of these matches at once.  It will perform up to 10,000 computer simulations that model the inheritance of DNA from the ancestral couple down to each of the DNA testers.  Then, it will compare the actual match data that I entered with the expected results from the simulations.

The first few lines of the output look like this.  For each match, BanyanDNA tells me the relationship in the tree, the actual amount of shared DNA, the average amount of DNA expected for the relationship, and a common range (±1 standard deviation for you math nerds).  It also tells me the number of standard deviations (SDs).

Simply put, the number of SDs is a measure of how far away the actual value is from the expected average.  Generally speaking, if a match is less than 1 SD from the expected value, there’s little to worry about.  If it’s 1–2 SDs from expected, it’s worth a closer look but not a huge concern.  And more than 2 SDs requires careful scrutiny before it can be accepted.

The screenshot above shows a subset of the results.  I had a total of 15 pairwise matches in this analysis.  Most of them were below 1 SD, so I am confident that my 1C1R is really my 1C1R and that these great grandparents were who I think they were.  However, my troublesome 2C1R, in the third row, is 2.7 SD from the expected average (actual: 19 cM, expected: 101 cM).  That’s not enough to say for sure that we aren’t really 2C1R, but it’s definitely worth a reappraisal.

The best way to reappraise is to throw more data at the problem.  In this case, I can also look at the match between my 1C1R (fourth row in the screenshot above, 399 cM and 0.3 SD to me) and Troublesome.  On paper, they are 1C2R to one another and share 160 cM.  That’s lower than the expected average of 213 cM but less than 1 SD.

Overall, this doesn’t definitively prove that Troublesome is my full 2C1R, but it gives me a little more confidence that they could simply be a low-matching cousin rather than an unexpected half relative.

Next Steps

BanyanDNA also supports hypothesis testing, so I can ask it to evaluate directly whether Troublesome is a full relative (one hypothesis) or half relative (a competing hypothesis) using Troublesome’s matches to myself and my 1C1R.  That’s a topic for a future post.  With only two matches to Troublesome, the results aren’t statistically meaningful yet.

Once AncestryDNA starts showing me how much DNA Troublesome shares with the other testers in the family, I can reassess whether Troublesome’s great grandparent was a half-sibling.  Time (and more data) will tell!