Chromosome Painting at Ancestry

AncestryDNA recently introduced a new feature that maps your ethnicities onto your individual chromosomes.  This beta chromosome painter can be found in the “DNA Story” tab.

SideView™

The chromosome painter builds on technology that Ancestry’s scientists introduced with SideView™.  SideView™ is a technological advance that can apportion your ethnicity estimates to the contributions from each parent.  It does this by comparing your DNA to that of your matches.

Think of it this way:  if both of your parents have tested, all the system has to do is compare your DNA to theirs to determine which copy of chromosome 1—we have two copies of each chromosome—came from Mom and which came from Dad.  The same with each of the other chromosomes.  Even if only one parent has tested, the chromosomes can be sorted by process of elimination.  This sorting process is called phasing.

But if neither parent has tested, things get trickier.  Consider a match who shares a single DNA segment of 25 cM with you.  The algorithm will be able to phase that small portion of your DNA into Parent 1 and Parent 2, but it won’t be able to tell which parent is which, nor will it have any insight into the rest of that chromosome or any of the others, for that matter.

Now consider a first cousin match, like the one shown here.  These two cousins share about 1,100 cM across 34 segments, with at least one segment on every chromosome except numbers 14 & 21 and the X chromosome.  The algorithm still doesn’t know whether this is a maternal cousin or a paternal cousin, but it can at least conclude that all 34 of those segments came from the same parent.  We’ll call this Parent 1.  By process of elimination, the complementary segments on each chromosome must have come from Parent 2.

Another first cousin on this side is likely to add an additional 600 cM or so to our Parent 1 phasing.  The algorithm knows the segments came from the same side because the two cousins will also match one another.  (This phasing algorithm can be misled by endogamy.)

More distant cousins can help, too, especially if they share segments on two or more chromosomes.  The algorithm will phase your DNA using cousins on both sides, sometimes assigning DNA regions based on a positive segment match and sometimes by elimination.

With a large enough database, the method should be able to phase all or almost all of your genome.  AncestryDNA is positioned to make the most of this technology, because they have the largest database of the genealogy companies, more than 22 million people as of June 2022.  Compare that with the next largest database, 23andMe, which has roughly 13 million genotyped customers.

Once the DNA is assigned to Parent 1 and Parent 2, the ethnicity estimates can be divided up accordingly.

 

Chromosome Painting

Which brings us back to the new chromosome painting feature.  All of the major genealogy DNA companies estimate ethnicity piece-by-piece along the chromosomes, and both 23andMe and FamilyTreeDNA offer chromosome paintings.  These are visual representations of where along your chromosomes each ethnicity is found.

AncestryDNA is now showing us the same information, but with the added benefit of sorting the chromosomes by Parent 1 and Parent 2.  Even if neither parent has tested there.

To see your chromosome painting, click the blue “View Breakdown” button below the SideView™ circle (see first image above), scroll down, and click “Explore now” in the Chromosome painter panel.  Yes, it’s a bit hidden; I hope they’ll make this feature easier to find once it’s out of the introductory beta phase.

Here’s mine:

It’s pretty neat!  You can see that one of my parents is mostly Germanic and Irish (teal and medium blue segments) while the other is mostly French (green).  This aligns well with the family histories of my German–Irish father and Cajun mother.

Those of you into DNA Painter will be glad to know that the inimitable Jonny Perl has already created a way to import the Ancestry ethnicity painting into his tools!

 

A Deeper Analysis

I do have some critiques, though.  The first is that my segment assignments from AncestryDNA and 23andMe aren’t particularly similar, as you can see.

For this comparison, I imported segments from both companies into separate DNA Painter chromosome maps and tweaked the colors to be similar.  In both maps, my paternal chromosome is above my maternal one.

At first glance, the most obvious difference is that AncestryDNA favors solid chromosomes of a single ethnicity while 23andMe shows more variation across most chromosomes.  This was a bit surprising given that AncestryDNA has a more fine-grained reference panel (77 regions) than 23andMe (44 regions), so I’d expect AncestryDNA to be able to give me more precision.

Of course, which company is more correct is an open question.  On this front, I’m inclined to believe 23andMe.  Why?  Consider my paternal chromosomes.  Only two of them (chromosomes 3 & 4) show any ethnic mixing.  My father’s mother was 100% German while my grandfather was about 60% Irish and 40% German.  It’s unlikely that the chromosomes I inherited from my father would be all one or the other.

What’s more, AncestryDNA shows chromosomes 9, 11, and 20 as all Scottish and chromosome 15 as all Sweden & Danish.  I have no known ancestry from either region, and any unknown heritage should be in segments much smaller than an entire chromosome.

Then there’s the Spanish segment on my maternal chromosome 8 (purple).

Roughly 2,000 Canary Islanders settled in southern Louisiana between 1778 and 1783, including some of my ancestors.  Although an imperfect proxy, my tree suggests I am about 5.5% Spanish.  Yet, Ancestry finds only one segment making up 1% of my genome.  By comparison, 23andMe estimates me at 9.7% Spanish & Portuguese.

What’s more, this is my mom’s chromosome 8 at AncestryDNA:

And here’s the Spanish & Portuguese estimate for me from 23andMe on chr 8:

The Spanish fragments just barely overlap!

I suspect the two issues are related, that Ancestry’s ethnicity algorithm penalizes a switch from one ethnicity to another, causing chromosomes to be assigned a uniform ethnicity when they shouldn’t be and impacting where the real breaks should occur.

 

The Good News

Fortunately, the chromosome painting is still a beta feature, meaning that AncestryDNA‘s scientists and programmers are still working on it.  And it’s a great addition!  AncestryDNA recently announced that their annual ethnicity update will happen in the coming weeks.  I am looking forward to seeing how the chromosome painting feature improves!

21 thoughts on “Chromosome Painting at Ancestry”

  1. Great post. BTW, Ancestry is going to do another update in a couple of weeks giving new regions, refining the regions of where our estimates come from and I believe will potentially update our ethnicity estimates again. I know you know that but a lot of posters might not.

    Secondly, I haven’t been to see my own chromosome painting. I’m glad that you explained the beta program because it gives me confidence to be patient for when they will roll this out for everyone and that it will be more refined.

    1. I’m looking forward to the update. I wanted to get this post out now to encourage people to compare before and after. I love seeing how the technology advances over time!

  2. Thank you for reviewing this new feature. I shared this with some of my students. We examined this feature today in class.

  3. This feature was a segway into a discussion about Chromosome painting, which for some, is an advanced feature.
    I liked the clarification and observations you made about Ancestry’s use of full chromosome matches on partial segments. It helps to know it is less accurate so you render the proper interpretation.

  4. Even with both parents tested, Ancestry has problems. I won’t go through chromosome-by-chromosome, but our daughter’s ancestry assignments don’t completely match up with either my wife’s or mine.

    Mostly, it’s a matter of calling a chromosome Irish for one and Scottish for the other; or England & Northwestern Europe for one and Germanic Europe for the other.

    In addition, my daughter is assigned two ancestries that Ancestry claims are from me, despite telling me I don’t have either. (But neither does my wife.)

    The “Germanic Europe” I definitely *do* have. In fact, it’s my largest single ancestry — and over a third of the total. Unfortunately, Ancestry lumps it all in with England & Northwestern Europe.

    Still, I think this tool has potential.

    1. I agree that it has potential but needs more work. I think (but don’t quote me on this … I don’t have the time to re-read their White Paper right now) that the whole-chromosome problem is happening because they’re working with probabilities. For each section of DNA, there’s a probability that it’s, say, Germanic and a probability that it’s England and NW Europe. And the probability may switch in an adjacent section of DNA. But probabilities are just probabilities, and they can be wrong. What I *think* they’re doing is saying, ‘Hey, this section is has a 75% chance of being German and the next one has a 51% chance of being English, and the third one is 51% German, so we’re going to call the whole thing German.” That’s a reasonable approach, but they’ve set the switch threshold too high.

      1. Yes, the fact that they’re dealing in probabilities is the reason they provide ranges for their assigned ancestries. Unfortunately they do *not* do this when the calculated score ends up being 0%. They actually did this once upon a time.

        For example, a person might click on “see all regions” and see that there was one or more regions which were not assigned — because the calculated percentage turned out to be 0 — but for which there was still a calculated range. Now they only give ranges when the final score is not 0%. This is too bad, because giving those ranges would allow customers to see ancestries that were at least considered.

        In my daughter’s case, the final score for her Germanic Europe was pretty small, only 2%. But the range was fairly large, from 0-23%. A range like that implies greater uncertainty in the assignment, and I strongly suspect that a similar-sized range is quite possible even if the final score is 0%.

        That’s why I think Ancestry should go back to showing ranges in such cases, at least for ranges beyond a certain interval. For example, where the difference between the high and low is greater than 10% or so.

        I also wish my experience with Ancestry were more similar to yours. I’ve also tested with 23andMe, Family Tree DNA, LivingDNA, and MyHeritage. Of these, I think Ancestry is actually the most distant from my well-established paper trail, with the exception of “Indigenous America”, where the estimate almost exactly matches 23andMe’s. (In fact, for this one ancestry, the two companies’ chromosome paintings are virtually identical.)

  5. The one thing encouraging that I have noticed about Ancestry’s chromosome painting is that my mother’s Ashkenazi Jewish segments correspond very closely with the Ashkenazi Jewish pile-up regions I have identified on FTDNA. I definitely agree about the bias towards assigning one ethnicity to an entire chromosome, though! Most bizarre is that Ancestry assigns two entire paternal chromosomes, one entire maternal chromosome to Norway, and one entire maternal chromosome to Sweden and Denmark, despite the fact that my paper trail (which is well supported by DNA matches) shows no Scandinavian ancestry at all. (And if some is hiding behind a brick wall, it certainly accounts for a very small percentage of my ancestry on either side.)

    1. I see a similar issue in mine. They assign me three complete Scottish chromosomes and one Sweden & Denmark. I have no known ancestors from either region. I do have Irish and German, though, which are geographically adjacent to those regions. I assume that’s the basis of those assignments.

      1. Interestingly, I think some of my “Scottish” (which greatly increased in the latest ethnicity update) is actually German, which they have always greatly underestimated. This is based on comparing Ancestry’s chromosome painting with the grandparents, great-grandparents, and 2G grandparents I’ve assigned to certain segments based on visual phasing. (I do have Scottish ancestry, but not as much as Ancestry is now showing.) I had been assuming the missing German was all or mostly being called Scandinavian, but it appears more complicated than that. When I have time, I’d like to systematically compare all of my VP results against Ancestry’s chromosome painting.

    1. That’s a great question! For myself, 23andMe is closest to my (verified) tree, with AncestryDNA a close second. I suspect which is best will vary by ethnicity.

  6. It’s very nice to see this new feature on Ancestry. I find the colours a challenge to identify. Perhaps they will incorporate some of the red and yellow family of colours into the mix. Or, have I missed seeing that I can change the colour for a region?

    Thanks for the write-up re this new feature.

    1. We can’t change the colors that I know off. I imported my ethnicity segments into DNA Painter and tweaked the colors there.

  7. I have a question – I did the testing through Ancestry, and according to the Ethnicity Inheritance graph (the donut), my parents have nearly identical proportions of Scottish (30% 27%), Irish (10% and 9%), and England/Northwest European (10% and 14%), which strikes me as highly unlikely given their origin stories. My mom has done a DNA test, and she’s definitely my mom, but maybe someone else is my dad? Or am I being needlessly weirded out??

    1. Do you have DNA matches who you know are related to you through your father? I’d focus on that rather than the ethnicity estimates, which are still a developing science.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.