Tuesday, January 25, 2011

Update on 23andMe's v3 results: Relative Finder comparison and ~30,000 v2 locations missing from v3

This will not be a scientific analysis, but, rather a preliminary summary of what I have found so far in my new 23andMe v3 results. My v3 Relative Finder matches loaded this morning and I have done a quick comparison to my v2 Relative Finder. There are some minor changes, but nothing that appears to be too significant.

(all are predicted)                         v3        vs       v2
3rd Cousins                                          4                     4
4th Cousins                                        38                   40
5th Cousins                                       262                 258
Distant                                                 33                   32
Public Match - 4th Cousins               3                     3
Public Match- 5th Cousins               10                   13

% DNA shared 
mother                                           50.00              50.42
sister                                               53.10              53.69
paternal uncle                               21.10              21.14
maternal aunt                                22.61              22.77
niece                                                24.31              24.49
1st cousin                                        11.45              11.36
1st cousin                                        11.95              11.94
1st cousin                                         8.38                8.37
1st cousin once removed               3.12                3.12
1st cousin twice removed              1.42                1.42
2nd cousin                                        1.15                1.15

Upon a brief initial review, predicted cousins lower than this appear to overwhelmingly share the same percentage as before.

Jim McMillan, who runs the independent "Cousins" matching project, compared my raw v2 data file to my raw v3 data file and this was his conclusion,

"There are at least 30,346 locations that are in v2 that are not included in v3.  So when a v2 is compared to a v3 there are going to be some longer half-identical segments that might not be there when comparing a v2 to a v2."

He also posted on his project thread on the 23andMe Forum:

"I have received v3 results from a contributor who also contributed v2 results. About 30,000 locations that appeared in v2 are missing from v3. Thus, when a v2 data set is compared with a v3 data set, it would be expected that the half-identical segments would be a little longer, and more segments meeting a 5cM and 700 SNP threshold found. All my software tolerates different numbers of SNPs. Where no SNP is found, it is deemed the SNPs would be at least half identical. On first blush, the missing SNPs do not appear to be evenly distributed among chromosomes 1-X.

In Chromosomes 1-x there are 30337 locations missing, and in Y 250, and in MT 172.

I have uploaded to box.net the locations in v2 that are missing in v3. See the details in the zipped spreadsheet at http://www.box.net/shared/b4shehl2qn."

I just received a Twitter message from 23andMe that the v2 SNPs were not removed intentionally. It explained (with some humor), "...some faction always falls out during chip manufacturing process. Biology is messy."

As I mentioned on my blog last night, my Ancestry Painting is still 100% European. However, from a review of the forums, the Ancestry Painting algorithm seems to be much improved for many.

Some customers are seeing improved updates to their haplogroup assignments, however there seems to be a problem with errors in the mtDNA mutations being reported from the new v.3 raw data.  They differ from the FMS results that some of us have received from FTDNA. I will post more information on this as it becomes available.

I am still working on getting details on the new SNPs. Hopefully 23andMe or one of the independent project admins will soon comment on the hundreds of thousands of new SNPs, since that is the most significant change and on what we should be focusing. I hear there is a Spittoon Blog Post on the way. More to come...

No comments:

Post a Comment