23andme

[DKF]

At least at the moment every European tested is not going to have say 6% Asian and African - this is a significant flaw in the decodeme offering - it almost guarantees minority ancestry - but entirely false. Any real ancestry of this type is either not there because of poor reference samples, or is embedded in so much noise it is virtually impossible to find the signal.

Sorry, what I meant to say is that every person of strictly European descent is going to test between about 2 and 8 percent Asian and African with decodeme. Their definitions and algorithm are too "liberal" or generous (loose?). This is an indication of the amount of noise that will confound the attempts of those with true minority Asian and/or African to pinpoint the correct percentage, and the specific location on the chromosome(s) where this ancestry is found. Since they have a very sophisticated browser, learning how to use it would be time well spent. If for example an individual has only one parent with minority ancestry, then seeing a block (e.g., the color yellow which represents Asian) of say two million base pairs and at the 45 or so percent for majority of the extent of the block (it will actually look like a block or an extended spike) - may indicate Asian, Native American, or Oceanic ancestry.

[burto]

If for example an individual has only one parent with minority ancestry, then seeing a block (e.g., the color yellow which represents Asian) of say two million base pairs and at the 45 or so percent for majority of the extent of the block (it will actually look like a block or an extended spike) - may indicate Asian, Native American, or Oceanic ancestry.

It's interesting you say that-on chromosome 5 of Mum's genome browser, when I set it to show ancestry, there is a heck of a peak...when I hone in on it, it's on one branch (i.e. from one parent) and it reads as being 65% Asian!!! There are a few other peaks on a few other chromosomes too. I'm still unsure of what the base pairs look like?
And when you say block, is that one branch of the 2 that make up the whole chromosome, or is it tiny parts of each branch?
According to DeCodeMe, Mum's 6% autosomal and X Asian matches are not significant, but these peaks I find interesting?

[burto]

Hello again,
Just had a look at that chromosome 5 and the bit that's 65% East Asian runs from about 129,000,000 to 132,000,000-is that what you mean by base pairs or is what I just wrote something completely different?
Also, does one half of the chromosomes represent one parent or are they mixed up?

[rainbow]

does one half of the chromosomes represent one parent or are they mixed up?

Is it in numerical order?
I don't have decodeme/23andme, but with my DNA Tribes I wondered if alleles in one column represented only one parent and the other column represented the other, but I noticed it is in numerical order (15,16. 11,12. etc) the lower number is first.
If yours isn't in order like that, then maybe one column does represent one parent. Try asking decodeme (oops, this is a 23andme thread) about that in an email if you want a solid & sure answer.

[DKF]

Hello again,
Just had a look at that chromosome 5 and the bit that's 65% East Asian runs from about 129,000,000 to 132,000,000-is that what you mean by base pairs or is what I just wrote something completely different?
Also, does one half of the chromosomes represent one parent or are they mixed up?

They have not been phased, where each nucleotide based is assigned to its proper chromosome. Hence it is all scrambled and by convention everything is set to the right. So you could have 58% and this might mean 50% from one parent and 8% from the other, but if could be 18% and 40%. So this is really not all that helpful unless the "wave" actually looks like a block, or the spike is extended over say 5 million base pairs. Still I have a 15 million block assigned to "African" but by a more thorough analysis, this is an illusion. Yes this sort of block appears among Africans, but also among Europeans but not East Asians - so it gets colored olive, and African it is (a striking error that I am sure has led many to erroneous conclusions).

There are a number of people who are working on a better way (more precise) to assign percentage of ancestry. If done correctly one can even "see" percentages as low as 1% which appear to be parsed out of the noise. It may be a while before decodeme and 23andme (despite my hounding) will offer this type of more effective analysis - but the data is there. I have had analyses using the programmes BEAGLE and PLINK. It is amazing what can be done - now if we can just convince the companies to up the priority on this, otherwise we will still have something that is only a cut above the 21 and 176 marker tests - there are half a million markers to work with. It is possible to be accurate to over 99%.

[burto]

They have not been phased, where each nucleotide based is assigned to its proper chromosome. Hence it is all scrambled and by convention everything is set to the right.

Sorry, I meant on the full genome scan, not the bit in the ancestry section, however that peak on chromosome 5 on the genome scan does appear on the ancestry chromosome page as a peak too. So what I mentioned about it being 65% East Asian running from about 129,000,000 to 132,000,000 was on the full genome view, if that makes sense?

I don't know what you mean by blocks and how I'd know if this peak was over 5 million base pairs?

[Maria_W]

Willing to share...

Willing to share my 23andme results with anyone who has tested with 23andme. Just send invitation to MariaW.

Maria

[Starr173]

All,

I have begun to agitate for improvements to the Ancestry Painting feature at
23andme. Nothing has changed since day one, which also means that they have yet to get around to painting the X chromosome. It is all static and will
remain so unless people step forward and express their opinions so that the
scientists can convince the resource allocators that it is imperative that
something be done to address this problem of they will lose business. I
have specifically not ordered testing of my uncle until I have direct
confirmation that 23andme understands that there are probably as many
customers interested in the ancestry component as the medical and trait
components (which are truly of no interest to me - the only thing I learned
that tweaked some curiosity was that I have a sprinter genotype).

When the present 23andme Ancestry Painting feature was designed only the
HapMap data was available to them - and to date there have been no
upgrades. Thus at the moment the Chinese and Japanese are proxies for all
East Asians, South Sea Islanders, and Native Americans. Their algorithm
appears to be quite good, such that there are unlikely to be many false
positives (a severe problem with decodeme data analysis). The problem is
that there are an unknown number of false negatives. My analysis (case
study) via Anders Palsen using the 52 worldwide populations via the
HGDP-CEPH database uses a much more robust approach to the search for
evidence of minority Native American. It succeeded in completely removing
any African on the X (decodeme had attributed 8% to this source) - there was nothing there, only evidence of a number of blocks shared by Africans and Europeans but not East Asians. Anders also found evidence of millions of bp matches to East Asians but only from the northeast, and to Native
Americans. He usedthe PLINK phasing program, and validating it by a direct
inspection with Excel formula. Hence my 100% European seen in the 23andme
Ancestry Painting is incorrect, and according to the analysis by Anders is
almost exactly what is expected based on the paper trail which includes
multiple sources.

Anders found 6.5% Native American (via Yakut and Xibo proxies) on the X, and a further 6.5% not found in any European sample and in the region flanking the above. This is likely a founder effect comparable (perhaps) to mtDNA haplogroup X2a found in the Great Lakes area.

The above brings up another point. At present there is not a representative
sample for the region where my ancestors lived, and so it is necessary to
employ the best proxies. Since I don't envision any change soon, at least
they could use northeastern Asian tribes instead of Chinese. I agree that
the approach they use will allow some percentage of the minority ancestry to
be caught in the net - but how much slips through?

I have started a thread at the 23andme community:
https://www.23andme.com/you/community/thread/741/.

Please use this or write to 23andme directly if you feel as I do that it is
high time that the company address the needs of those of us interested in
autosomal and X ancestry. If they do not receive a flood of "constructive
criticism" they will come to the conclusion that we are all happy as clams
and put their efforts into beefing up other parts of their product. Only a
concerted effort by all of us will work. I know that the scientists are
willing and able, but the company needs to know that there is an interest or
"demand" out there for improving and upgrading the Ancestry Painting.
Thanks.

I agree completely with you, Dr. Faux. I left an e-mail with chiah about just offering the ancestry painting as many of us are not interested in our risks for various medical disorders. Of all the genetic tests out there I thought 23andme was the best one but I'm beginning to doubt that it may be understimating minority ancestry. Is there any way that PLINK program can be used with DNAPrint results? I haven't tested with any of the other companies because I wanted to make sure how accurate they are before I spent my money. I was indeed thinking of doing the 23andme because of the lesser price. I tried to leave a message at the link you provided but I don't think I can post because I only have a demo account there.

[DKF]

I agree completely with you, Dr. Faux. I left an e-mail with chiah about just offering the ancestry painting as many of us are not interested in our risks for various medical disorders. Of all the genetic tests out there I thought 23andme was the best one but I'm beginning to doubt that it may be understimating minority ancestry. Is there any way that PLINK program can be used with DNAPrint results? I haven't tested with any of the other companies because I wanted to make sure how accurate they are before I spent my money. I was indeed thinking of doing the 23andme because of the lesser price. I tried to leave a message at the link you provided but I don't think I can post because I only have a demo account there.

At the moment 23andme is our best bet. I have spoken with the scientists. They clearly "get it". What is needed is for the managers to authorize this work. I really believe that the company is convinced that we are all jazzed about the medical and trait features. I am in the biz and all I can say is that I am not impressed. This is not the fault of 23andme, it is just that the hopes about genetic markers and their link to common diseases has not panned out at all well, and probably won't until full genome scans are offered (in 5 years?). Our best hope is to make a pest of ourselves, asking specifically for features that are of interest to us. There is no way that the average individual is going to be able to use PLINK or BEAGLE to do the proper in depth analyses unless Anders and Doug decide to go commercial.

[tomcat]

An interesting discussion on 23andMe markers, Native American ancestry, and a guy that is offering a BGA analysis of 23andMe results, starting here:

http://archiver.rootsweb.ancestry.com/th/read/GENEALOGY-DNA/2009-04/1239372692

[Noaide]

I haven't tested with any of the other companies because I wanted to make sure how accurate they are before I spent my money.

The chips itself provide wealth of data, however neither decodeme and especially 23andme have not been able (yet) to exploit the full potential of the data from yourself and the reference populations.

[Maria_W]

Shut down...

I invited so many people to share their genome with mine that I shut it down. Can't access it or compare to anyone else....

[rucksack]

I invited so many people to share their genome with mine that I shut it down. Can't access it or compare to anyone else....

Keep all of your "partners".

It works for less sharing, but they are working on a fix.

It might be a couple of weeks.

I'm not going to toss out my, so far, 181 partners just to compare less. I'm happy to wait for a fix.

[brian87]

ordering from 23ANDME

Hello,

I was about to order the 23andme test but I wanted to make sure of some things first.

Is the company stable?

they are not about to go under like dnaprint?

How long will the results really take at this time?

Are they about to upgrade or lower their price? Should I wait a little while?

I am mainly interested in their autosomal admixture results but I wonder if I can use it to narrow my FTDNA HVR1 and HVR2 mtDNA matches?

Thanks,
Brian

[tomcat]

Hello,

I was about to order the 23andme test but I wanted to make sure of some things first.

Is the company stable?

they are not about to go under like dnaprint?

How long will the results really take at this time?

Are they about to upgrade or lower their price? Should I wait a little while?

I am mainly interested in their autosomal admixture results but I wonder if I can use it to narrow my FTDNA HVR1 and HVR2 mtDNA matches?

Thanks,
Brian

They seem reasonably successful and are not under the same financial cloud as the Icelandic DeCodeMe.
Some folks have gotten results within the 10 week period but like all tests delivery time is an unknown.
No one but 23andMe can say what their future prices may be.
Depending on your MtDNA haplogroup, 23andMe's Mt SNP's may be of some use in refining your place in Mt phylogeny but will not be as definitive as FGS. Their Y SNP's are also useful.