Saturday, September 26, 2009
Miscellaneous This's & That's
From the Genealogy-DNA Lists:
By Patrick Tagert -
"Re: [DNA] Relative Finder Beta Testing at 23andMe
What is the first & foremost lesson that we can learn from the new feature?? It struck me as transparently obvious as I stared at the host of unidentified "4th, 5th,6th & distant cousins". More than 95% of our potential relatives are unseen in the current version of 23andMe. Without this new Relative Finder feature, any given customer is unlikely to benefit from knowing even a small fraction of their potential "cousins". If you have an interest in 23andMe for genealogical explorations of autosomal DNA, it was reasonable to refrain from any investment in their products, prior to this revision. It remains to be seen how contact can be made with people who potentially share common ancestry within relatively recent generations. This hasn't been announced or reviewed as part of the Beta version, & the Beta is an experiment (at least to some degree) to determine the public's interest in this aspect of a service that was originally conceived as primarily focused on medical/health risk assessment. So is now the time to buy the new product?
Just a couple of points. From now to the end of September, the Full Version of the 23andMe test is available at a very reduced promotional discount price. If you have been holding back for the above reason, or any other reason, you probably won't get this opportunity again in the near future. The reduction is more than the typical "holiday sales". It is an "unpublished" promotion, in the sense that it is only being promoted by Beta testers, who are free to describe it to some degree on public blogs, so long as price specifics aren't included in the discussion lists & forums. Remember, the this new feature is a Beta, so the feature may change before it is finalized, but the current FULL VERSION that is being offered will not change, & that's what you will be getting with this limited time promotion. There are five days left to order. You can order the FULL VERSION using a discount code from any of the folks who have posted to this discussion with details of the Beta product review. All of the Beta testers have discount codes that will be they will be happy to provide to you if you want to order in the next five days. The promotion ends September 30.
By Roberta Estes -
"Part of the problem I've had so far is that we've done a very good job of educating our groups about Yline and male descendants and mito and maternally descended people. It took a long time to get there but they do seem to understand. Now we're switching gears on them and they're having trouble understanding that this has no bearing on the line of descent or who takes the test (male, female or how related/descended)."
By Ann Turner -
"There have been lots of interesting questions and observations. I'm not ignoring them, but I'm not clear on what my role should be here. I am certainly not an official representative of 23andMe, and I'm not privy to some of the ins and outs of the Relative Finder algorithm. 23andMe is in the midst of moving to new office space, so that complicates things, too. However, I have been thinking about this sort of thing for quite some time in connection with my own study of the familial hearing impairment, even before Relative Finder was introduced. I will consolidate some questions, post a few thoughts, and hope they are not worse than useless.
I'm assuming some basic background here -- for those who haven't had occasion to even think about autosomal DNA yet, I recommend SMGF's tutorial on different modes of inheritance:
Autosomal SNPs have two alleles, one inherited from the father and the other from the mother. It is not possible to tell which is which; the alleles are always given in alphabetical order. If two people share at least one allele, they are considered to be half-identical. A long stretch of consecutive SNPs that are at least half-identical doesn't occur just by chance: it's evidence of a common ancestor in the not-too-distant past. In other words, it's identical by descent (IBD), in contrast to identical by state (IBS, which can be just a coincidence). Leon Kull coined the abbreviation HIR (he pronounces it "her") for half-identical region. A HIR which is long enough shows up on the Family Inheritance diagram at 23andMe.
1) 23andMe's method for estimating a relationship takes into account the number of consecutive SNPs that are at least half-identical, AND the length of the segment in cM, AND the number of segments. Sharing two segments on different chromosomes that add up to say 12 cM is a stronger clue of recent ancestry than one 12 cM segment. The cM (centiMorgan) is a unit for "genetic distance" (a different concept than genetic distance for Y-STRs), and it reflects the recombination rate, which varies in different chromosomal regions. If two people are half-identical for a region that rarely gets split up by a cross-over point, that's not as meaningful as identity for a more active region. Once the boundaries of a HIR are established, I believe they are graphed according to the "physical distance" (the position of the SNP in bases on the chromosome).
2) The chances of two people inheriting the same region from a common ancestor diminishes rapidly with each generation, as there's only a 25% chance of passing it on in both lines (50% x 50% in each line). However, if the region does survive this lottery, it may remain lengthy enough to be recognizable for many generations. The chance of a cross-over point occurring smack dab in the middle of the region is rather low. Thus you may not have an IBD segment with a known 4th cousin, but all of your DNA came from SOMEwhere, and you may connect through one of your many ancestors to one of another person's many ancestors in the more distant past.
3) The threshold for the Family Inheritance diagram was 10 cM (about 10,000,000 bases ON THE AVERAGE). Relative Finder has set a lower threshold, but if it's set too low, you get too much "noise" -- a European can even find a HIR with a Bantu.
I've put together a spreadsheet based on my prior reading I think it might answer some questions, and it has a link to a file with the European / Bantu example.
Please take note of the fact that I've spent quite a bit of my own time assembling this response, and I need to return to my major assignment as a consultant for 23andMe, which is locating and annotating citations for the Health and Trait reports."