Saturday, April 4, 2015

Autosomal DNA A tale of two stacks of pancakes

Basically, I like pancakes. So I think about autosomal matches and their segments like stacks of pancakes. Everyone gets two kinds of pancakes. One from your mom and one from your dad. Even if you don't know which one is which, you can actually start sorting them into those two sides.

I'm using 23 and me here because it has the most straight forward chromosome matching system I've found outside of Gedmatch.com.

So here's an example of a bunch of overlapping segments from genetic matches. This is from my dad's perspective. See chromosome 1 there with 4 little pancakes stacked up. He matches Thomas, Jeremy, Paula and Jenette in roughly the same area. Some pancakes are bigger and some smaller, but they roughly stack up.


That is a mixed stack of pancakes. We'll call them blueberry and strawberry. We don't know which are pancakes from mom's side and which are from dad's side or if all of them are from the same side.

For now we'll consider that they might be blueberry and strawberry pancakes mixed together.


The goal is to sort them into two matching teams. We want to know all the strawberries and all the blueberries and only compare them to each other.

Why? Well, we're figuring out which side of the family to focus on. When you're looking at family trees if you can narrow the results in any way, you can cut thousands of possible relatives out of the search, just by splitting your matches into sides.


From my dad's perspective we would have to search each family tree for each of four people for all the ancestors of each person back six or more generations. From my dad's perspective they could all be related to the same person. But what if they're not? Everyone matches my dad, but do they all match each other and therefore all share the same ancestor?

The way you find that out is by sorting them based on how they match each other. In genetic genealogy you need to see the world through someone else's eyes. I need to know which pancakes are blueberry and which are strawberry..if any.

So one way or another, I need to become one of the other matches for a moment and see how they see this same stack of pancakes. 23 and me and Gedmatch make that easy..you can just check people against each other to see how they line up. FTDNA comes close with chromosome browser and the matrix...Ancestry folks are on their own (they can't do any of this).

Begin sorting.

Let's become Thomas then. How does Thomas see this same stack of pancakes. The trick is, Thomas will only see his matches. He is either a blueberry or a strawberry pancake and he will only see his kind of pancakes in the stack.


Aha! There he is. Thomas matches Thompson of course..we knew that, but Thomas also matches Paula in the same spot. We can arbitrarily label them as strawberry pancakes. Of this stack of pancakes they only match each other and my dad.


Next step. Pick one of those 0 people who were missing. They "should" match each other, but we need to confirm it. Does Jeremy match Jenette?


Goal. Jeremy matches Thompson (as expected) but only sees his own kind of pancake, Jenette. We now have a blueberry stack as well.


Keeping in mind that I have no idea which side of my dad's family they are on..I do know that they are on one side or another. Further, I will only look for common ancestors between Jeremy and Jenette on the one side and Thomas and Paula on the other.

In this way, your segments (pancakes) can validate each other. They form "triangulated groups". People who match you and each other in the same area on a chromosome, thereby sharing a common ancestor among all the matches.

What if you go through this and there are no blueberries and only strawberry pancakes? Well, you'll just have to wait for some blueberries to show up. What if you find a pancake that doesn't match any of the other pancakes, blueberry or strawberry. That is a pancake that is probably an error. Someone who matches you by accident. Typically larger segments (pancakes) are less likely to be one of those accidental matches.

Just like with regular genealogy, genetic genealogy is skeptical. You have to PROVE IT. You can have hunches and that is okay, but it's not enough that two people who are genetically related have the same person in their tree.

You need at least a third person with that same ancestor to match you on the same chromosome, roughly on the same segment (a little bigger or smaller is okay). Just like you need more than two legs in a stool, you need at least three legs in a genetic match. Figuring out which legs belong on which stool is all a sorting game and you can do it before you ever look at a family tree.