Archiver > GENEALOGY-DNA > 2008-02 > 1203878218

Subject: Re: [DNA] 23ANDme versus deCODEme
Date: Sun, 24 Feb 2008 13:36:58 EST

Now that 23andMe and deCODEme both have demo records, you can explore the
user interface aspects yourself. Both companies are introducing new features
lickety split, though, so it's a moving target.

I don't think you should make a decision based on mtDNA or Y features, unless
you're only interested in "deep ancestry." Neither company can provide the
resolution of the genealogical HVR/STR tests. However, 23andMe does get down to
much more detail in the haplogroup labels. 23andMe is using a smaller number
of Y-SNPs (current reports make use of 284, vs 858), but these seem to be more
targeted toward the currently defined phylogenetic tree. We don't know yet how
the additional Y-SNPs in the deCODEme set will pan out, and deCODEme is not
reporting subclades as deeply as they could right now (i.e. just R1a and R1b).
I suspect lots of the extra SNPs will end up being "phylogenetically
equivalent" for the deepest splits in the tree, like the long lists of SNPs in that
paper by Underhill and Kivisild, but we'll only know that when we've seen actual
results for many people in different haplogroups.

Both companies are making use of the HapMap samples for geographic ancestry
in the 50-some population cell lines from the Human Genome Diversity Project,
and in fact 23andMe partially funded the analysis which is the basis for the
current flurry of technical articles. They must be using a slightly different
algorithm in their presentations, though -- deCODEme says it breaks the
chromosomes down into segments for comparison purposes, and that seems to change the
rank order slightly. Only deCODEme gives an admixture estimate, and that only
for the three broadest categories of European, African, and Asian. You can,
however, compare your results to a representative sample from any of the 50
populations at deCODEme on a segment by segment basis.

23andMe has a nice Genome Browser in place, where you can use the web
interface to look up individual genes or rs numbers you encounter in the technical
literature (or their own Spittoon blog, which has frequent posts about items
that crop up in news releases). DeCODEme doesn't have a genome browser yet, but
promises it will be "soon." In the meantime, you can of course check your own
alleles in the big download set and use dbSNP to learn more about the gene.

As for the medical aspects, I think you should thoroughly explore the sample
reports and understand what's really being offered. They are association
studies using "common" SNPs, not diagnostic tests for single mutation / single gene
disorders, like the ones DNATraits is now offering (with a special emphasis
on mutations more commonly found in people with an Ashkenazi Jewish

23andMe's Gene Profiles strike me as chattier, with historical background and
interviews with specialists, but both companies cover the fundamentals. Both
companies also do an admirable job of providing citations to the literature,
with data about the frequency of the genotype in cases and controls, and how
that contributes to relative risk. Many pundits are complaining that people
might get freaked out by learning that their risk is doubled (the way newspapers
often report new findings), but I have to wonder if they've examined the actual
reports, which provide more details about the absolute frequency etc. Most
people will learn that there are multiple genes "associated" with each
condition, with some alleles adding to the calculated risk and others subtracting from
it, and a doubling of risk might take you from 2% to 4%.

It's very important to realize that these are population studies, and the
statistical associations don't necessarily apply to an individual case. If you
have a "high" relative risk for colon cancer, the odds may still be low in
absolute terms that you will contract it. If you have a low relative risk, that
does not absolve you from the need for regular examinations, since not all causes
of colon cancer are known yet (and may be idiosyncratic for a particular
family line, like the George Frye pedigree we discussed recently). That kind of
mutation might never be picked up on a chip using common SNPs.

The SNPs aren't necessarily causal, either, even the small percentage found
in actual coding regions. In fact, most of them are probably SNPs in the
vicinity of the mutation responsible for the actual effect, and they're just along
for the ride as that segment of the chromosome is passed down through the
generations. They are tagging/mapping SNPs that provide guidance for which genes to
study in more detail. That's the main business of deCODE, and they are
folding their discoveries into the deCODEme reports as soon as they are published.
Of course, 23andMe has access to those articles, too.

The feature I found most intriguing at deCODEme was the graphic display,
comparing any two gene profiles on a segment by segment basis (about a million
bases). You can do a similar thing at 23andMe, but in practice, it only
highlights connections for first degree relatives, who share broader regions of their
chromosomes. But now that both companies give access to raw data, anyone can
create a novel approach to visualizing data (sort of like open source software)
-- if people are willing to share their profiles. You don't need to share data
about individual SNPs for the deCODEme display, like the ones I've uploaded

One remaining difference between the companies is the number of SNPs, and
deCODEme has a head start there. That may prove to be the deciding factor for me,
as I have a scheme in the back of my mind for locating the mutation
responsible for the hereditary hearing loss that runs in my family. We've already
participated in several genetic studies, but the mutation has not been found. I
could test various relatives (for instance the two sons of my first cousin) to
see what segments I share with the just one who has the hearing loss. The sticky
points are deriving haplotypes from genotypes and downplaying segments that
we share because of our generic European ancestry. I'll need to do some pretty
heavy-duty cost-benefit analysis here!

On the other hand, 23andMe has added some custom SNPs to their suite. There
are autosomal SNPs, in addition to the ones for mtDNA and Y, but I haven't seen
a comprehensive list of why they were selected. One example is a test for the
CCR5-delta32 mutation (HIV resistance), which some of you have ordered as a
stand-alone test.

Neither company is really promoting the genealogical possibilities for the
autosomal SNPs. They may be (justifiably) pessimistic about that application,
since there's only a 50% chance of passing on a particular DNA segment to the
next generation. Nonetheless, all of our DNA came from somewhere, and we now
have unprecedented access to huge amounts of data. We ought to be able to mine a
few little nuggets out of it, and I'm lobbying the companies to pay attention
to this segment of the market. I may at some point enter into a consulting
arrangement with one or both companies, at which point I'll make a full
disclosure on this list. Or I may not feel it's proper to do that at all, since my aim
is to remain objective while representing the interests of the genetic
genealogy community.

In the meantime, I think the tests are most suitable for those willing to
explore the next frontier, with all its unknowns and with the possibility of less
expensive tests coming online within the next few years. The cost-benefit
analysis will be tricky for everyone, not just for my own little niche. But if
you decide to proceed, I don't think you can go wrong with either company.
There's a third company, SeqWright, which is flying under the radar right now, due
to their "late" entry into the field.

Ann Turner

In a message dated 2/23/2008 4:33:42 PM Pacific Standard Time,

> Ann and List:
> I am still wrestling with which company to employ in a genome scan of 1
> million or so SNPs to shed light on ancestry, physical traits, medical
> conditions (risk assessment) and so on.
> Never being one to reinvent the wheel, I will ask on the eve of my decision
> for some input as to whether at this point one company's offering stands out
> above that of the other in any of the above categories. Clearly a grand is
> a fair bit of coin, so the choice must be made wisely (particularly so since
> retirement is looming and an opportunity to fund this sort of endeavor may
> not come along again for a while).
> Whether on-List of off-List I would very much appreciate an assessment
> (even very brief) of the merits of turning to one company or the other. Now that
> 23andMe has a Y-chromosome component to their testing, they are very much in
> the game as far as I am concerned. However they seem to measure only about
> 2/3 the number of raw SNPs as deCODEme - which is certainly a consideration /
> concern.
> Thanks for any help the pioneers can offer to me as I weight the blessings
> of each option.

Ideas to please picky eaters. Watch video on AOL Living.


This thread: