PDA

View Full Version : 23andme Updated Ancestry Painting



Scarlet Ibis
11-14-2012, 06:01 AM
Just making a thread now so that people can share their experiences, and opinions with the updated ancestry painting when it rolls out. As far as I understand from reading the 23andme message forum, some users are on the beta version of it, but they're asking those users not to reveal themselves or give public feedback yet. The launch date for everyone should be coming up soon.

Humanist
11-14-2012, 07:14 AM
Just making a thread now so that people can share their experiences, and opinions with the updated ancestry painting when it rolls out. As far as I understand from reading the 23andme message forum, some users are on the beta version of it, but they're asking those users not to reveal themselves or give public feedback yet. The launch date for everyone should be coming up soon.

Will certainly post my results when the time comes. Not expecting much. Hopefully they surprise me.

AJL
11-14-2012, 04:37 PM
I notice they've changed from the navy blue from "Europe" to "Europe/Near Eastern" if you click on My Ancestry.

I am wondering whether my grandfather, who had a little Asian in his painting and lost it a couple of years ago when they rejigged things, will regain Asian, and whether my mother, who has a slight amount of Asian, will lose hers.

AppalachianGumbo
11-14-2012, 11:22 PM
From what I understand, they are not *redoing* your genotype that provided your percentage as European, African or Asian. They are only going to break out populations from those groups in which you placed a percentage or percentages to give an idea of what populations you are similar to in either of those 3 categories.

For example, if you currently have no Asian in your 23andMe Ancestry Painting, then you will not see any populations associated with the Asian populations, (i.e. Chinese, Native American etc.) in the AP2. Likewise, if you have no African in your Ancestry Painting, you will not see any African populations.

If you are 100% European, those will be your populations. I think some people are under the impression, that if they are 100% European, they may show some Asian or African with AP2.

YGG
11-16-2012, 01:27 AM
From what I understand, they are not *redoing* your genotype that provided your percentage as European, African or Asian. They are only going to break out populations from those groups in which you placed a percentage or percentages to give an idea of what populations you are similar to in either of those 3 categories.

For example, if you currently have no Asian in your 23andMe Ancestry Painting, then you will not see any populations associated with the Asian populations, (i.e. Chinese, Native American etc.) in the AP2. Likewise, if you have no African in your Ancestry Painting, you will not see any African populations.

If you are 100% European, those will be your populations. I think some people are under the impression, that if they are 100% European, they may show some Asian or African with AP2.

This is not accurate. 23andMe are phasing our data for AP2, so it is possible that the regional percentages will differ from AP1. In fact, some will definitely see changes in this regard.

AppalachianGumbo
11-16-2012, 08:31 PM
This is not accurate. 23andMe are phasing our data for AP2, so it is possible that the regional percentages will differ from AP1. In fact, some will definitely see changes in this regard.

Yes, accurate. Also, *Phasing* has nothing to do with what 23andme is doing with AP2, assigning chromosomes into ancestral populations or groups. For example, Native American is categorized as Asian with 23andMe. People with Asian attributed to Native American may find more of it or most of it identified as Native American proper. However, most people 100% Euro, the wonderful deep blue are not going to see drastic or big changes carrying over into other racial groups such as African or Asian if it is typically not seen now in your painting. They will see their Euro broken down into European populations. The population are added to further deepen your European, African or Asian result. They have added Native American but will still be included in the Asian group.

People are in for big heartbreaks if they think some big changes are going to be seen. However, if it makes you feel better, yeah...okay.

AJL
11-17-2012, 04:48 AM
Phasing can change how ancestry is read rather drastically.

As an experiment, if you have DNA for a child and both parents, try creating two homozygous versions of "child" chromatids, one duplicating the father's contribution and the second the mother's. I did this and surprisingly the ancestral paintings of both were quite a bit different not only to the actual child, but also to either parent.

AppalachianGumbo
11-17-2012, 03:50 PM
Phasing can change how ancestry is read rather drastically.

As an experiment, if you have DNA for a child and both parents, try creating two homozygous versions of "child" chromatids, one duplicating the father's contribution and the second the mother's. I did this and surprisingly the ancestral paintings of both were quite a bit different not only to the actual child, but also to either parent.

While I can agree with this, I have seen some of the beta of the AP2, most Europeans have no added African or Asian ancestry if not already in their paintings now. We will see. I just think some people have high hopes there is going to be big changes. I can't get into other details, but people will see.

GTC
11-17-2012, 04:01 PM
Well, as I understand it, there cannot be any phasing of autosomal data unless you have also had both of your parents tested.

I wonder what percentage of 23andMe customers fit into that category?

Then there's phasing of the X chromosome, but I don't know much about that.

Human
11-20-2012, 11:48 AM
Well, as I understand it, there cannot be any phasing of autosomal data unless you have also had both of your parents tested.

I wonder what percentage of 23andMe customers fit into that category?

Then there's phasing of the X chromosome, but I don't know much about that.

If you have enough "cousins" (IBD)... maybe it could be accomplished without parent data. The problem here would be the IBS "cousins, of course, but with a high enough number of "cousins" at a higher block size...hmm...
Are there any other limiting factors I am forgetting here?

jowston
11-20-2012, 02:44 PM
Well, as I understand it, there cannot be any phasing of autosomal data unless you have also had both of your parents tested.

You should be able to phase your data with one parent tested. You won't be able to recreate the missing parent's data, but you should be able to determine in most instances the source of the markers.

Ann Turner
11-20-2012, 03:54 PM
Well, as I understand it, there cannot be any phasing of autosomal data unless you have also had both of your parents tested.

Parents and children are not only easier to compute -- they are still the gold standard for phasing (assigning the two alleles in a genotype to the proper parent).

But it's possible to do some phasing with population data. What follows is my very superficial understanding of the method. It's a statistical process, where you compare one person's genotypes to a number of genotypes found in a population sample (or even better, to a dictionary of haplotypes). You then try to assess the probability that a particular allele in one SNP is traveling with a particular allele in an adjacent SNP based on genotypes in your reference samples. You make a few random guesses and pick the best match.

You have to do this for "sliding windows" of a limited number of SNPs at a time, since the number of possible combinations increases exponentially with each SNP. This can sometimes result in "phase switch errors" -- you think you are traveling along the maternal chromosome, but all of a sudden you jump to the paternal chromosome. However, you get decent results much of the time for stretches of useful length.

I was scouting the web for some good background material, but I haven't found exactly what I want yet. I get lost pretty quickly in this presentation, but it's a start.

http://dna.engr.uconn.edu/bibtexmngr/upload/Pas08.ppt

thetick
11-21-2012, 02:57 AM
http://dna.engr.uconn.edu/bibtexmngr/upload/Pas08.ppt

As always a great response Ann, thanks for posting here. I think you did a good job explaining the complexity of phasing, not just a simple child to parent comparison. That presentation has lots of goodies. I just glanced at it and will need to spend quite a bit time to understand it. I've dealt with the math before (entropy and probability calculations etc) but not in a genetic context, so understanding this population phasing will be a fun and hopefully successful challenge.

Although Ann if you getting quickly lost, I'm not sure anyone here will have a better understanding than your description above. I think almost all us here are like myself just knowledgeable amateurs in the genetic genealogy community.

Scarlet Ibis
12-02-2012, 08:55 PM
An update from 23andme member staffmember "epistate," that I got from the 23andme homepage feed:


It looks like Ancestry Painting’s successor will launch next Thursday, Dec 6th, provided a few final technical checks pass. Note that, when Ancestry Painting’s successor launches, the original Ancestry Painting feature will no longer be available.

ChrisR
12-03-2012, 01:01 PM
An update from 23andme member staffmember "epistate," that I got from the 23andme homepage feed:
Looking forward for this. Phasing would be also a perfect feature; let's see if FTDNA or 23andMe made it available first.

Yorkie
12-11-2012, 09:12 PM
I quite like the updated Ancestry Painting. Aside from the obvious British/Irish major slice of my ancestry [to be accurate, it is largely English], the new version nicely picks up a slice of French/German and Scandinavian. I guess my 17% unassigned Northern European also contains a fair bit of continental Germanic and Celtic. Solving a family argument, I got absolutely zero Italian.

GTC
12-12-2012, 09:00 AM
Here's what they make of me:

86.5% British and Irish
2.6% Scandinavian
1.5% French and German
9.1% Nonspecific Northern European
0.2% Ashkenazi
0.1% Nonspecific European
0.1% Unassigned

The 9.1% "Nonspecific Northern European" has me intrigued.

I, among many, have asked the various BGA tool creators how far back, in terms of generations, these signals can be read. I have never received a satisfactory answer and I suspect that the people who produce these tools don't really know. My male ancestor supposedly went to Ireland c1169, having reportedly been Anglo-Norman before that. I guess after some 840 years (say 28 generations) my signals are strongly British and Irish as shown. However, I don't know much about my maternal line other than it's Irish and English back to about the 1800's and then the trail goes dark.

History-of-Things
12-12-2012, 04:11 PM
Aha. Well. My results:

Conservative Estimate:
98.9% European
.2% British and Irish
55.4% Nonspecific Northern European
.1% Iberian
<.1% Southern European
43.2% Nonspecific European
1.1% Unassigned

Standard Estimate:
100% European
9.6% British and Irish
72.8% Nonspecific Northern European
.2%Iberian
<.1% Nonspecific Southern European
17.4% Nonspecific European

Speculative Estimate:
100% European
37.4% British and Irish
18.8% French and German
1.6% Scandinavian
36.8% Nonspecific Northern European
.3% Iberian
.8% Nonspecific Southern European
4.3% Nonspecific European

It would appear than my genes have "left the Old-World building" and therefore are largely nonspecific. Still, it beats the heck out of the Genographic Project's attempt at such things. It is interesting that the continental elements are almost equally split by my mother and father's side, when I would have expected my mom's German roots to be more pronounced than my father's Huguenots. The Scandinavian is on my mother's side (no surprise--New Sweden bits and bobs), as is the Iberian--there lies a mystery, which may in fact have something to do with my mtDNA lineage....