Affy to Illumina conversions complete?

[nathanm]

I have a current business project where we finally decided we needed three algorithms to evaluate the final result - sparse data, rich data and in between. It feels like the algorithm for FF should be the same. Too many matches and you do not know where to start. Too few and you wish the algorithm was looser. They are all cousins, some more distant.

I had some known relationships based on the Affymetrix results - 6th cousins and 8th cousins for example. The 6th cousin match gave us a major way around a dead end and eventually linked to a paper trail. The 8th proved a paper trail too. These were real matches and the algorithm for the Illumina seems to have sent them away. I think we may need those options. For me having a couple hundred distant cousins is interesting and useful, for someone else it is not. For someone else having even a few matches is the issue.

John

I like the idea from your business. A one-size-fits-all approach doesn't seem to be working so hot for FF. It's possible there was a bug like gtc posted above; if so I hope they fix it quickly.

Your example of tracing the paper trail to distant cousins is exactly what I was asking about before. If they don't want to adjust the criteria for matches, the least they could do is provide a way to manually enter known relationships that aren't showing up. It shouldn't be that hard: enter the two kit numbers and known relationship. FTDNA could then compare both FF results internally, to verify there's some shared DNA, just not enough to meet the normal match threshold.

[dwight]

I don't think any of us here would disagree with this statement in the least!

Therefore a greatly increased database size could really help us.

[JPHutchins]

So maybe we should complain less and try to recruit more people to take the Family Finder test.

All that said, I still think their communications department needs a major overhaul!

Carol Anne

I agree on both counts. I joined the DNA revolution to find a way thrugh brick walls, too. Unfortunately for me and other adoptees, DNA is often the only hammer available to break through the wall. Therefore a greatly increased database size could really help us.

[dwight]

Interestingly enough, *if* the news in the post by "gtc" 3 messages prior to yours turns out to be true, then it would seem that our "complaining" (and i would characterize most of it as simply asking for clarity, looking for explanations, and, yes, expressing some frustration with the lack of either) actually paid off in stirring the FTDNA team into examining their processes and finding a glitch - a glitch that may explain the rather large portion of Affy matches many of us were reporting as AWOL...

It occurs to me that if we keep complaining on this bulletin board the way we always do, we're going to chase away the people we need to make the database grow so that it will be more useful to us! I'd like to have the Family Finder database a hundred times it's present size.

I did the test hoping to break down some brick walls. No luck yet!!! BUT between this company and 23andMe, I match 3 different people descended from my 7th great-grandparents. That tells me the test actually works! It also tells me that my research was correct, that none of the documents in my paperwork were wrong (forged, for instance), and that there were no non-paternity events in my direct line (or in those of the people I matched). Pretty good!

I may be wrong, but I'm not sure 23andMe is all that interested in developing the genealogy side of it's operations. I think they are more interested in the health aspects of DNA. I'm hoping for major improvements in tools for interpreting autosomal results from FTDNA because genealogy IS their business.

So maybe we should complain less and try to recruit more people to take the Family Finder test.

All that said, I still think their communications department needs a major overhaul!

Carol Anne

[MoberlyDrake]

Complaining!

It occurs to me that if we keep complaining on this bulletin board the way we always do, we're going to chase away the people we need to make the database grow so that it will be more useful to us! I'd like to have the Family Finder database a hundred times it's present size.

I did the test hoping to break down some brick walls. No luck yet!!! BUT between this company and 23andMe, I match 3 different people descended from my 7th great-grandparents. That tells me the test actually works! It also tells me that my research was correct, that none of the documents in my paperwork were wrong (forged, for instance), and that there were no non-paternity events in my direct line (or in those of the people I matched). Pretty good!

I may be wrong, but I'm not sure 23andMe is all that interested in developing the genealogy side of it's operations. I think they are more interested in the health aspects of DNA. I'm hoping for major improvements in tools for interpreting autosomal results from FTDNA because genealogy IS their business.

So maybe we should complain less and try to recruit more people to take the Family Finder test.

All that said, I still think their communications department needs a major overhaul!

Carol Anne

[JohnG]

Missing Matches/Too Many Matches

I have a current business project where we finally decided we needed three algorithms to evaluate the final result - sparse data, rich data and in between. It feels like the algorithm for FF should be the same. Too many matches and you do not know where to start. Too few and you wish the algorithm was looser. They are all cousins, some more distant.

I had some known relationships based on the Affymetrix results - 6th cousins and 8th cousins for example. The 6th cousin match gave us a major way around a dead end and eventually linked to a paper trail. The 8th proved a paper trail too. These were real matches and the algorithm for the Illumina seems to have sent them away. I think we may need those options. For me having a couple hundred distant cousins is interesting and useful, for someone else it is not. For someone else having even a few matches is the issue.

John

[Darkwriter]

Need to do more

I ordered the FF and mT testing in Oct. hoping to beat the big chip-switch. Unfortunately, FTDNA sat on my request until Feb. to utilize the new chip, which I too have evidence of being less than accurate. Realizing most groups and individuals all used the same method for testing, I had my autosomal results run through DNA T****S SNP test to narrow down the regions and groups I am more closely affiliated with (genetically.) I found this analysis to be much more helpful to my paper trail hunting because those odd admixtures can be seen and their relative regions more carefully researched for ancestry. There is no use in having more FF/autosomal tests rerun if all the groups discussed here are using the same chips and/or methods: insanity has been defined as repeating the same action but expecting different results.
We do have to keep searching for different tools and methods to extract that which we cannot find, but know exists. Nothing is perfect, but don't stop reaching for other means of testing, paper trailing and creating those GEDCOM files to check against each others'. Now, that is the biggest time consumer I have encountered so far, but I suspect my own missing ancestors will finally be found that way.

[gtc]

A little bird tells me that after a project admin conferred with FTDNA about lack of matches, FTDNA discovered that there is a software bug with the new Family Finder tests which results in kits being compared to only half of the database. Apparently a fix is being implemented today.

I have no more information other than that, so let's see what eventuates ...

[MoberlyDrake]

I've lost more than half of my matches.

One thing we have to give them credit for. If Illumina is really more accurate, they did at least upgrade everyone at their expense! And, if I understand correctly, the Family Finder test is still in the beta phase. Doesn't that mean we're the "guinea pigs" and they have the right to change or scrap things as they think best?

But how hard is it to send out mass emailings informing their customers about what is going on (especially the beta testers, if that is what they consider us)? Other companies do it all the time. I think I get email from Ancestry.com almost every day, to say nothing of any online business I happen to have ordered from! One person should be able to compose the message and send it out to all customers at very little expense, one would think.

Carol Anne

[gtc]

it would be nice if the "softer" support services were seen as just as necessary to the overall mission and customer satisfaction.

Fully agree. Has FTDNA ever commissioned a professional customer satisfaction survey covering all aspects of its service including communication?

[gtc]

Recent graduates are encouraged to apply!

Yep, as I suspected: lowest rank on the totem pole.

[Kasandra]

I've lost all but five of my Affy matches. As many of you will remember, I was upset and contacted FTDNA this year when cousins I had been working with here for MONTHS didn't show up as matches on 23andMe. I had a few experts look at the raw DNA and FTDNA had reported an area with a decent sized mismatch in the middle of it as one solid piece. I spoke to Mr. Greenspan about it and he, and his son, insisted that the person I was specifically enquiring about, was indeed my cousin. They said they 'ignore' mismatches in that area. IF that's the case, then on what basis dare they remove the AFFY chip cousins? I paid them to find me cousins, it's called FAMILY FINDER, not TEMPORARY FAMILY FINDER.

It is my suspicion, that these cousins that have not translated to cousins on the Illumina chip are not cousins at all. Or only very distant ones, and that 'compensating for' the error rate on the Affy chip has led to creating too wide a window and allowing too much chaffe in with the wheat. (and I will remind everyone that I have said this was the case and worried about this for months-long before they decided to convert to Illumina). Further I think that they are burying these results under the rug as quickly as they can in the hopes we'll forget or not have the reach to affect their future sales because the Affy chip was just so flawed that the relations are shotgun results. That being the case, I feel that Mr. Greenspan lied to me on the phone, which I take very personally.

The thing is, if it was just a mistake, they should come out and say it. "We didn't realize how big a difference that error rate would make. We apologize to all our customers." There how hard is that? Would I be pissed. Oh yeah. Would I get over it? Yep. Will I get over the feeling that I can't TRUST them now? No. Trust is EVERYTHING in this game. OMG, the number of hours I've spent... The number of people I've inconvenienced and the time they've spent pouring over their family trees, looking in old boxes for information, contacting other people in their family...it's just crushing. I specifically mentioned to Mr. Greenspan that I was very concerned that one day I would approach someone and say, "Hi, you may be my brother, can I get a cheek swab?" based on the information he was providing and that if he said he was certain, I would trust him. Now what do I think?

[Javelin]

On the Affy conversion here's another twist. On DNA-forums there are some saying they have not seen their FF results converted yet. Doug McDonald is one. He has an Affy kit and no Illumina replacement for it.

Me neither. My guess is a small percentage has not yet been processed, I don't know, maybe 2-3%.

[mkdexter]

Thanks for the links and info Paul.

Yes FTDNA would be very wise to move forward on the RF - FF import.

On the Affy conversion here's another twist. On DNA-forums there are some saying they have not seen their FF results converted yet. Doug McDonald is one. He has an Affy kit and no Illumina replacement for it.


mcdonald, on 27 April 2011 - 10:30 AM, said:
really?

I turned in my request at the very first and still no results.
I did have to send in another scraping, but they reported my full
111 markers results promptly. Something funny is going on.

Doug McDonald


Are all Affy conversions really done? Very interesting.

[Geneadict]

Alex Khomenko, a 23andMe software engineer who just left 23andMe, announced last Nov 2010 that they had 60,000 customers. I think newer numbers have been estimates from customers. I don't believe there's a way to see how many of the 60,000 are just there for the medical test but its a significant number of them.

I estimate FTDNA is at 9500 FF tests. I have not asked them because I don't really care anymore. Paul usually does his homework though.

The database growth is more important once you're in the system. It used to be 1000 a month give or take for both companies but its higher now and it fluctuates depending on sale periods etc.

Same source, noted 75,000 as of Feb 2011 in these articles (which were posted in a 23andMe community thread by Glenn Nolen)

http://www.genomesunzipped.org/2011/...s.php#comments

http://www.quora.com/How-many-paying...s-23andMe-have

Based on new matches, 23andMe has also clearly had higher ongoing growth rate since December 2010 as of result of lower pricing and various sales thus the reason for my earlier estimate of 80-90k.

(also, 50k threshold was announced late June 2010)

http://spittoon.23andme.com/2010/06/22/23andme-and-a-new-paradigm-for-research/

On the FTDNA side, my estimates may still be on the high end. I originally thought FF was close to the 10k mark last summer after their "Bull Rush" BETA orders and public launch and comparing match ratios to 23andMe's Relative Finder (RF). However, Matt set me straight and it turned out that my Affy match ratios were nearly double that of my father-in-laws and others and thus part of my large match loss on FFs Illumina.

Of course both numbers are moving and you may get different answers depending on who you ask and whether you are talking about kits issued/sold versus results in the database (as there would be a lag) and in the case of 23andme would also be a small percent that have opted out of RF included in the totals.

Of course for 23andMe there are the issues with low response rate and health vs genealogy interest that have been addressed in numerous places in the past year along with other Pros/Cons for both 23andME and FTDNA's FF.

Only other thing that I can say is that I haven't had a materially higher rate of identifying actual relationships for one versus the other, but have had a couple more closer matches pop up in 23andMe, which you would expect, having a higher liklihood of such with the larger database. Of course that also comes down to the random luck of recombination and who tests with which company and even more random luck of whether you can contact them or not. May also depend on an individuals circumstances with regard to which test would be better, so if they can do both, that would probably be a fairly consensus recomendation. If FTDNA eventually offers an import option for 23andMe data for FF (at a nominal fee of course), the best course may be to purchase test with 23andme and then do the data import. Of course that will also depend on the pricing and individual needs.