Thursday, October 29, 2009
23andMe - FYI
Subject: [haploblock] from DNA forums
I work at 23andme -- I read this thread and thought I would add a few points.
- Family Inheritance uses an older algorithm than Relative Finder, so the size of identical chunk found by relative finder can be smaller than family inheritance. I believe the two methods will be unified.
- As has been pointed out, the algorithm finds small, usually single, (essentially) 100% identical chunks. This is very different from Compare Genes, which is a general similarity measure. The two do not necessarily correlate. Relative Finder is the more sophisticated, relevant measure, I think.
- The reason that the size of the chunk can be greater for more distant cousins is that the size of the chunk is not as important as the "genetic distance". Basically, the important factor is the typical number of recombinations in this location. As I think has been pointed out, regions near the centromere often pop up as "blue" in family inheritance -- because the number of recombinations is low in this region.
- We are still working on the best way to do sharing etc. I don't know much about this, but it is still beta!
- The false positive rate should be very low for this method. The reason is that the chunks we detect are (essentially) 100% identical. However, the concept of cousin is a tricky one. A single chunk can pass down through many lineages unaltered just by chance, which can make it look like your tenth cousin (or even twentieth cousin if you are unlucky) is your sixth cousin. The resolution is limited because after you go out past fourth or fifth cousin, you usually share just one chunk. This is why we always include a range. This limitation can be overcome using data from parents, grandparents, etc.
- We are very excited that the genetic genealogy community has picked up on the power of this tool so fast. I think we'll see a lot of useful information coming out of it.