Regarding AncestryDNA and one's child

Okay. I have an ancestry DNA profile (hypothetically) as follows:

Europe West 50%
Great Britain 26%
Ireland/Scotland/Wales 20%

The rest are low confidence. Let’s assume that we do a test and my daughter’s percentage of Europe West DNA is 27.7%. Does that mean that I can count the 25% she got from me, subtract it from her total: 27.7-25=2.2 and then double that and say with any confidence that her mother is 4.4% Europe West?

And then do the same with the others? Or does it just not work like that at all?

It’s essentially random. You get 50% of your DNA from each parent, but that doesn’t mean you get an equal portion of each subset of DNA. So in your hypothetical example, the daughter got 50% of your DNA but maybe she only got 10% of the West European markers. Or maybe all 27.7% of her markers came from you and 0% came from the mother.

Imagine you make a pizza, but sprinkle the pepperonis randomly. Your friend asks for half of the pizza. Does that mean he will definitely get half the pepperonis? No. First we have to know which slices have more pepperoni, then we have to know which slices you actually gave him. He will definitely get half a pizza, but we can’t predict whether he gets half the pepperoni. Repeat the trial enough times and he will probably average out to half the pepperonis, but in any one trial his pepperoni would be more or less than exactly half.

There are other problems with these tests. For example, if your test says you have Western European DNA, that doesn’t mean your DNA definitely came from Western Europe. It just means your DNA resembles what people from Western Europe often have. Additionally, each analysis agency has their own “library” of samples, so the results are different depending on which agency is doing the comparison.

My mom had a large percentage of DNA from Scandinavia. Mine was less than 1%. I didn’t get how this was possible but I guess if it’s like a lottery that makes sense.

To clarify, Europe West DNA only means that is who you group with from that testing company.

They take a very small, specific selection of Autosomal DNA, that happens to repeat and they have a “chip” that when they unzip the DNA, they can count those repeated segments. So a tiny percentage, of non-randomly selected DNA groups close to other customers that happen to self report from being from that area or perhaps ancient DNA which happened to be found in that area.

It is not really that meaningful outside of the entertainment value.

The math gets crazy quick, and yes I have even paid for full mtDNA and Y sequencing but it is still very hard to get useful information from.

Autosomal DNA can get “washed” out or weird segments can carry forward. And the numbers of relatives grow by an exponential amount to the point that if you have any European ancestry you are a direct descendant of any person who was alive in the year ~900 AD who has a living ancestor.

This doesn’t mean that you can’t gain information from the test but that the value is far lower, especially in the presented form than they will tell you.

Using similar techniques I would be a very close relative to Ust’-Ishim by the methods they use as an example, and even if I am related, unless time travel was involved I am not that close.

For the most part consider it an amusing approximation.

Interesting. So when the test estimates that my second cousins are my second cousins, is that due to the “pepperoni” effect, or is that looking at, for lack of a better term, the cheese, the anchovies and the black olives as well? Is that test more accurate than the ethnicity test?

To add the math, the chip that uses, counts one small feature (repeats) for ~0.02% of your entire DNA. Now some of this is targeted, but those numbers also relate to percentages they can assign to other customers, and not the “percentage” of DNA that is anything.

We don’t know how much they think that they can assign, but 50% of ~0.02% is at most ~0.01% of the DNA, and that is assuming their chip picks STRs that are useful for ancestry and that there were no missed calls or wrong calls.

DNA is inherited in chunks, often get all of a chunk of DNA from a parent, or none of it. Seldom do you receive exactly half of a chunk, or ancestral segment but on average you will get half from a parent.

To clarify, 6 generations ago you had 2^6 or 64 *-grandparents at that level, and on average you would get 1.56% but you may get more or less or none at all.

You do get inherit 50% of each parents’ DNA which my last reply may have suggested isn’t true (typically).

mtDNA and Y are the exceptions here as those follow the matrilineal and patrilineal lines.

As one link above said

Is there an accepted standard deviation for this? Ideally, I would get 50% from each parent. What are the chances I get 75% from one parent and 25% from another parent.

And, not that I am worried about it at all, but just for information, is there something that would show up in these tests that would show with a pretty high probability that she is/not is my biological child?

I tried to clarify, you will get 50% from your mother and 50% from your father, but lets say your father is 50% “a region” you may get 0% DNA from that “a region” or 50% of “a region” but on average (meaning population wide) you will get 25 “a region”

But no, there is no way to standardize that, but the nice thing is that it doesn’t change your lineage.

What they do have that is fairly standard is a centimorgan which is basically an estimation on the rate of changes of the chunks you do get.

There is enough information to establish paternity, but ancestry especially in the case of DNA is not nearly as concrete as people believe.

See this page

It offers a good explanation of how the real world will differ from the mathematical models, along with a real nice chart to show the expected amount of shared DNA vs the range of actual amounts for most relationships.

What’s a “large percentage”? If it’s more than 51%, then it doesn’t make sense that you would have only 1%. Either the test results are wrong (either hers, or yours, or both), the test results are correct but have a huge margin of error (check the fine print), or the woman whom you call your mother did not actually give birth to you.

The genealogical DNA testing companies all test SNPs, not STRs.

Ethinicity estimates are, as you say, a lot less precise than the companies make you think, though. Except if you get unexpected results well above 10% one should not start drawing conclusions from them, and even then it ought to either be very unexpected, like a different continent, or well above 10% and not a neighbouring region.

As rat avatar explains, the test should look for long sequences that match perfectly between you and the possible cousin. The statistics of chromosome crossover during meiosis are consistent enough, I guess, that the number and length of these repeated sequences provide a fairly accurate indication of consanguinity.

Make a free account at If Ancestry has your DNA you can download it to your computer and then upload to Gedmatch, which provides access to a much broader range of analyses than Ancestry. Looking at a chart right now, I see “58 centiMorgans total” for a match with my 3rd cousin, and about 115 centiMorgans for 2nd cousins 1X removed, 1920 cM for a niece.

From there you can click to look at the details of the segment matches! I see my 123 cM match is 24,000+ “SNP’s” on chromosome 4. (Wouldn’t nucleotide be a better term than SNP in this context?) There is no constant ratio between cM and nucleotides: different parts of the genome have different crossover statistics.

Gedmatch accounts all have an e-mail address, which you can query to hope for genealogical information. (This is all in addition to various analyses available for ethnic background.)

~ ~ ~ ~ ~

BTW, I think there’s a simple explanation for some discrepancies between ethnicity tests. “Pure” West Europeans will have, depending on how far back you look, a mixture of East European DNA, etc. So different analyses may structure the report in different ways.

The tests can’t tell what DNA you got from which parent. This can give false matches for relationship matches, which look at sequences of matching SNPs, as it’s theoretically possible that what they think is a matching sequence actually zig-zags between your paternal and maternal chromosomes (and/or zig-zags in your match’s data).

The false matches are minimized by only looking at sequences that are at a minimum several hundred SNPs long.

Ethnicity matching on the other hand is based entirely on SNP frequency, so the error of margin is indeed huge. And when comparing mother and daughter they are completely irrelevant because the more precise sequence matching will tell, without a doubt, whether the two samples match.

Not a biologist, but… to clarify: Genetics 101

You have 23 specific distinct PAIRS of chromosomes, which contain (almost) all the DNA that distinguishes you. One “pair” is XY if you are male, XX if female.
You got half of each pair from mom and half from dad.
These are specific and diferent - pair 1 is different from pair 2 is different from pair 3 etc. So you don’t get (for example) 3 of mom’s pair 3 and one of dad’s pair 4.
When making eggs or sperm for Junior, your gonads create those egg or sperm cells with half a load - one of the two from each pair of chromosomes.

So to pick a pair at random - say the 3rd pair - if they are both “West Europe” in a parent then Junior will definitely get a “West Europe” as part of their 3rd pair, plus whatever the mate contributes from their 3rd pair.
If the parent has a WE and a Brit as pair 3, then Junior will get either the WE or Brit. Pick only one. (Plus whatever the other parent contributes…)
So simplistically, if one chromosome controls red hair, Junior has a 50-50 chance of having red hair. If mama has a pair where both have red-hair chromosomes, Junior will definitely get a red-hair one.

Except -
Chromosomes duplicate when cells split, including when they produce eggs and sperm; during the process, the DNA spiral “unzips” like a zipper and collects matching molecules on the two pieces to create two DNA spirals. During this “copy” mistakes can happen - wrong piece, or the pieces “break” and fix and so part of one of the pair may attach to the other of the pair, or duplicate chunks may occur, or any of a dozen possible problems that create differences (transcription errors) in DNA can occur. 99% of the time this change is irrelevant, or dangerous to the point of reducing survivability, but sometimes it creates a better specimen of the species.

Two points -
This mixing (transcription error) of the two chromosomes can mean that the chromosome you pass on may not be 100% from one parent, but could actually be a mix.
The harmless or useful changes can accumulate over time creating differences that spread among the population over time - which is how different populations have somewhat different genes.
The rate these chromosomes change is also roughly regular, giving an approximate clock on how long it’s been since two people had a common ancestor.

Plus as mentioned, mitochondrial DNA - little DNA rings floating around in the cell but not in the nucleus - are only inherited from the mother via the egg, as sperm cells don’t have them. However, when the cell splits there’s no guarantee you have a copy of every different one left in the gel inside that cell - you may only get some…

So you can tell who is Thomas Jefferson’s male line descendant (or from his brother or uncle) based on the similar Y chromosome; you can tell who has the same all-female lineage based on mitochondria.

So the math for the first generation is simple - you got half your chromosomes from each parent. Beyond there, it gets fuzzy. of your dad’s 23 chromosomes, some came from his dad, some from his mom. Like flipping a coin 23 times, odds are close to 50-50; but in specific instances, anything can happen (plus that possibility of a transcription error meaning not completely). The only certainty is that with 23, it won’t be exactly 50-50. Go back enough generations and there’s a good chance you have nothing in common with perhaps one of your ancestors.

So the odds that all the genes you pass on to your offspring are from only one of your parents - well, the odds of flipping a coin 23 times and getting all heads is 2^23 or about 1 in 8.3 million.

How does your comment square with this comment:?

Do I get 50% from each parent or possibly an imbalance? If it is indeed 1 in 8.3 million, I’m sure we can find an example in the world where a child has absolutely 100% of one parent’s genes and zero from the other.

The second comment you quote ignores Chromosomal crossover during meiosis, or rather it assumes the number of crossover events is zero. I dimly recall reading that there are, on average, about 3 crossover events per chromosome. We’d need detailed statistics on crossover to address your question, but getting no genetic material from one of your grandparents requires zero crossovers on all chromosomes during that meiosis and is HUGELY smaller than 1 in 8 million.

The categories are so broad as to be meaningless. Why would you expect any such precision from manipulating them in any way, and why would anyone wish to do so?

You absolutely get 50% from each parent (ignoring mitocondrial DNA 100% from mom, and the size difference between the X and Y-chromosome).

You do not, however, get 25% from each grandparent. That is theoretically possible to get entirely from just one. You’d have to have a gamete with zero recombination on any chromosome though, and the odds of that are in the order of 1:10*19, give or take a handful of orders of magnitude. And if that wasn’t a big enough hurdle, you then have to have all the whole chromosomes sort themselves into a grandpa-gamete and a grandma-gamete (It doesn’t matter which one of these you get for the purposes of this exercise). Which is where the 1 in 8.3 million comes back in.