View Full Version : 23andMe's v4 chip

11-21-2013, 08:36 PM

They're now using 600k SNPs instead of 900k

What implications does this have for 23andMe users in the future who are interested in exploring their ancestry through third party tools such as GEDmatch, or group projects like Harappa?

11-24-2013, 02:52 AM
Cost saving measure on their part?

Ann Turner
11-24-2013, 02:12 PM
We can't say for sure until we (and the third parties) see the SNP selection. It sounds like they are maintaining a core list of Ancestry Informative Markers, though.

GEDmatch is already able to handle records with different SNP sets for matching purposes by equating a missing SNP with a no-call. No-calls don't interrupt the long consecutive run of SNPs in a half-identical region.

FTDNA may not want to accept transfers, since they wouldn't accept v2, but their new IT department may be handle this like GEDmatch does. The reasoning is that if you match on a decent number of SNPs in a row, you would also match on any SNPs located in between ("linkage disequilibrium").

11-24-2013, 04:27 PM
They're now using 600k SNPs instead of 900k
According to the article, they will engage in the dangerous practice of imputation--i.e., guessing important alleles based on what Utahans and Sardinians have at that location.
The custom chip will also allow our researchers to better impute many variants present on larger chips. Imputation is a powerful statistical method that allows researchers to fill in missing data points. Researchers use publicly available full-sequence data, 23andMe’s own data and other sources to impute the genotype at specific locations.

Publicly available sequence data is a euphemism for absurdly biased datasets like that of the 1000 Genomes Project, which does not recognize the existence of modern humans anywhere between the Netherlands and China (nor in the Balkans or Middle East) (http://www.1000genomes.org/faq/which-populations-are-part-your-study).

Imputation is not totally unreasonable for the mere hobby of ancestry estimation; but when determining the presence of medically significant alleles, imputation is deceitful and unethical.

I myself was the victim of imputation: A couple decades ago, a blood bank "diagnosed" me with a loathsome and fatal disease based on an imputational test which, experts later admitted, had a false positive rate of more than 50%. Luckily, I immediately recognized their "diagnosis" as quackery.

Scarlet Ibis
11-24-2013, 07:25 PM
Just in case, I would suggest everyone download their raw data now before the changes take place. If they go back and delete the SNPs they're not going to be using anymore, your raw data might be worthless for Gedmatch, etc. I doubt they'd bother, but they might to save server space. Who knows.

I'm not sure it matters for me, since I'm still on the V2 chip, but I'm going to download my family's raw data anyway.

Scarlet Ibis
12-02-2013, 04:15 AM
John Olson, from Gedmatch, posted on 23andme that Gedmatch will likely make adjustments to accommodate the fewer SNPs on V4.

@Christine M,

You are correct that GEDmatch will likely make adjustments to accommodate V4 results.

It is my understanding that 23andMe Genealogical comparison results use SNPs that are essentially in common to both V2 and V3 chips. Will this trend for genealogical comparisons continue with the introduction of the V4 chip? It seems to me that most of the additional SNPs in the V3 chip are being used to identify health related traits. They also make a big difference in the GEDmatch eye color prediction tool results :-) Will we see this trend continue with the V4 chip? That is, will the V4 SNP changes be primarily related to assessing health issues?


John Olson