Affy to Illumina conversions complete?

[mkdexter]

I meant small segments that match between two people. Sorry .

[JohnG]

Why matches changed?

I am still looking for understanding on why there was such a large and systematic change in matches in the move from Affymetrix to Illumina.

1) We have more SNPs considered, so perhaps there are more chances for segments to not match (more chance of small gaps of non-matching SNPs).

2) There might be new segments considered or old ones not considered. One would think this would come out even though. If there are new areas some of the old Affy non-matches might have matched and you would get new matches that were Affy. (I do not know how to tell the difference between a new match from the old Affy clients and a new match from the new Illumina clients!)

3) The algorithm used for matching may have changed, either due to differences in the chips (error rates for example) or due to corrections for closely related populations, such as Ashkenazi Jews.

If it is (1), losing half your matches sort of shakes faith in the process. If we did 1 million SNPs would half my remaining matches go away?

I sort of hope for (3) and maybe the correction part of that. I lost a number of Ashkenazi, Eastern European, and Rhineland matches, and others that seemed to be New England or New Amsterdam colonial collections with founder effects causing lots of matches.

If it is the effect of a correction to reduce large numbers of 5th and distant matches, I would like an option to reset my matches to the old algorithm!

[gtc]

I am still looking for understanding on why there was such a large and systematic change in matches in the move from Affymetrix to Illumina.

Yes, we keenly await an official explanation from Max Blankfeld.

[varlene]

How many Affy matches are people now missing?

30 Affy Matches (All Speculative)
-2 Converted to Illumina
28 Missing = 93.33%

[Ann Turner]

Mother:
25 Affy.......1 converted to illumina
Total of 7 matches.....6 new ones.

Me:
49 Affy.......10 converted to illumina
Total 21 matches....11 new ones

My mother has lost about 95% of her Affy matches and now has less than 1/3 total matches than before the conversion. Can anyone beat that?

Were you the one who didn't inherit any of her mother's Affy matches? How about her Illumina matches?

[Ann Turner]

I'd be interested in comparing match lists for father/mother/child trios who have data for both the Affy and Illumina chips. Be sure to set your Matches to All before downloading the CSV files. My e-mail address is [email protected].

I have some stats from one family trio. I'd be interested in looking at still more non-Ashkenazi families. The Ashkenazi analysis is difficult because FTDNA has been removing some distant matches for them.

Discordant SNPs (e.g. Father is GG, Mother is AA, child is GG): genotyping error or microdeletions

Affy 1372/548011 = 0.00250
Illumina 207/708092 = .00029

Match in child not found in either parent (potentially false positives; need phased data to confirm)

Affy 48/121 = 0.397 [three cases were "fixed" in Illumina]
Illumina 14/106 = 0.132

Conversion

57/121 Affy matches not found in Illumina list
55 of those were in the "speculative" range (4th or 5th to distant)
43 of those are cases where the match in the child was not found in either parent

Median length of longest segment

converted 12.48 cM (range 7.74 to 46.04)
not converted 8.13 cM (range 7.72 to 17.72)

[LSSmith]

Were you the one who didn't inherit any of her mother's Affy matches? How about her Illumina matches?

I was the one who did not share any affy matches with my mother. I don't share any of her 7 illumina matches either. Dr McDonald has declared her 100% Lithuanian, this is an under represented population for testing. However, I'm still not sure why we can't find one in common. My cousin's results are in and none of us shares a match. My brother's results should be completed in June and it will be interesting to see what happens with his results.

[LSSmith]

I am still looking for understanding on why there was such a large and systematic change in matches in the move from Affymetrix to Illumina.


there might be new segments considered or old ones not considered. One would think this would come out even though. If there are new areas some of the old Affy non-matches might have matched and you would get new matches that were Affy. (I do not know how to tell the difference between a new match from the old Affy clients and a new match from the new Illumina clients!)


If it is the effect of a correction to reduce large numbers of 5th and distant matches, I would like an option to reset my matches to the old algorithm!

This is exactly what confuses me. I understand how you lose some Affy matches. However, I assumed that you would pick up some new matches as well. Thus in the end you would end up with about the same number of matches. I don't understand why me, my husband and mother have far fewer total matches from Illumina never mind them being old Affy matches or not.

TOTAL MATCHES
Me: 49 Affy to 21 Illumina
Mother: 25 Affy to 7 Illumina
Husband: 43 Affy to 30 Illumina

[canaryislandgirl]

No matches with mother

To LSSmith,

How can you have no matches with your mother ? I don't understand that. Unless you were adopted and not told or switched in the hospital at birth... Has anyone else had this same experience with family finder ? That doesn't give me much confidence in the accuracy of the results if this is really your birth mother. People who are adopted are relying on these results to be accurate. I am not adopted, but I can see how upsetting this would be.

[LSSmith]

To LSSmith,

How can you have no matches with your mother ? I don't understand that. Unless you were adopted and not told or switched in the hospital at birth... Has anyone else had this same experience with family finder ? That doesn't give me much confidence in the accuracy of the results if this is really your birth mother. People who are adopted are relying on these results to be accurate. I am not adopted, but I can see how upsetting this would be.

I have to correct my statement. I do share a match with my mother however, it was a match I paid for not a new find. We both match my cousin. My cousin wasn't tested under Affy. Without my cousin's DNA we do not share any matches and yes it was a bit upsetting and very disappointing. I want you to know that the testing does indeed show that my mother is in fact my mother! We match we just don't share other matches. At first I questioned the results and thought they must be wrong. Thankfully a few people looked at them for me and they looked correct. Due to the small size of the of Lithuanian DNA database and the fact that all of her matches are very distant I just don't share any matches with her. They are just too watered down when they get to me. (not a scientific explanation) However, all of my matches are distant as well and I am only half Lithuanian. My father's side isn't giving me any strong matches either. You can't fault FTDNA as they can't control who tests with them and who doesn't. So I am here waiting and hoping to get a strong match someday. I will match my brother when his results are in but I don't need testing to find him. Since we haven't any good matches from FF, Y DNA and MtDNA my brother thinks we are actually aliens.

[Ann Turner]

I was the one who did not share any affy matches with my mother. I don't share any of her 7 illumina matches either. Dr McDonald has declared her 100% Lithuanian, this is an under represented population for testing. However, I'm still not sure why we can't find one in common. My cousin's results are in and none of us shares a match. My brother's results should be completed in June and it will be interesting to see what happens with his results.

The "expected" result would be to share half of your mother's matches, but since she only has 7 total, and they are all distant, perhaps there's a threshold effect. Her matches might be just slightly over the threshold and yours just slightly under. If any of her matches are interested in pursuing this, I'd be glad to take a look at the raw data.

[nolnacsj]

30 Affy Matches (All Speculative)
-2 Converted to Illumina
28 Missing = 93.33%

Varlene, your results are even worse than mine! Only 7 out of my 33 Affys converted to Illumina, I am missing 26, but for you to only have 2 Affy's convert over to Illumina? Wow!

[econnore]

Deed is Done?

Well, if it's May 2011, I guess the Affy --> Illumina deed is done. I know that FF doesn't consider mtDNA results, but with the conversion of my 55 Affy results to 38 Illumina, I lost a few and gained a few. Interestingly, I lost 2 Affy matches that were also HVR1/HVR2 mtDNA matches.

[Geneadict]

This is exactly what confuses me. I understand how you lose some Affy matches. However, I assumed that you would pick up some new matches as well. Thus in the end you would end up with about the same number of matches. I don't understand why me, my husband and mother have far fewer total matches from Illumina never mind them being old Affy matches or not.

TOTAL MATCHES
Me: 49 Affy to 21 Illumina
Mother: 25 Affy to 7 Illumina
Husband: 43 Affy to 30 Illumina

If you examine your Affy matches for the location of the primary shared segment (most likely over 7 cM in length) and compare to your Illumina matches, I think you will find no Illumina matches with the primary segment in locations where the missing Affys had them. I had several such clusters in Affy that are gone in Illumina and thus net reduction from 87 to 36, whereas my father-in-law did not have nearly as many of these clusters and went from 85 to 86, but still loss of 35 Affys offset by 36 New Illuminas. So much of it comes down to the random luck of recombination and the makeup of the database.

[Javelin]

Well, if it's May 2011, I guess the Affy --> Illumina deed is done.

No, it's not. There's still at the very least that group of 100 kits not converted.