Change in Platforms:
On February 1st, 2011, Family Tree DNA changed testing platforms for our Family Finder autosomal DNA test.
We changed to the Illumina Omni platform. As part of that change, we retested for FREE all Family Tree DNA customers who previously ordered and paid for a Family Finder test. The new microarray chip for our Family Finder test is the Illumina OmniExpress. It tests about 710,000 unique autosomal DNA SNPs (single nucleotide polymorphisms).
The Omni platform provides better long-term upgrade options than the Affymetrix platform. http://www.familytreedna.com/faq/answers.aspx?id=39
Whoops, accidentally changed the number of SNP's code while trying out what worked... In any case, the one used from 2011 to 2013 was http://snpedia.com/index.php/Category:On_chip_FTDNA2 - it says "10,716 pages total" right now, but that could be little on the upper side. I can't figure if there's a way to invoke that count directly in text, though. Maybe 14,730/14,730 would work. In any case since 2014 they have new chip Illumina HumanOmniExpress-24 http://support.illumina.com/array/array_kits/humanomniexpress-24_beadchip_kit/documentation.ilmn (Not sure if this should be FTDNA3) with slightly different SNP list. I have 9,345 annotated SNP's on current run with it, but that of course depends a bit on number of no-calls etc. --Donwulff (talk) 05:12, 30 May 2014 (UTC)
That's the same array they upgraded to in 2011. The SNP list change is likely due to a change in reporting or scrubbing. Also the page is technically incorrect about Y-DNA SNP testing. The OmniExpress array does test a limited number of Y-DNA SNPs but FTDNA excludes those from the raw data they report. -- Jlick (talk) 16:23, 31 May 2014 (UTC)
It's not exactly same array, they switched from Illumina HumanOmniExpress-12 to Illumina HumanOmniExpress-24 together with 23andMe. (23andMe, of course, still using custom version with more health SNP's). Best source found is description at http://forums.familytreedna.com/showthread.php?t=33864 including the quote from Dr. Mittelman, FTDNA's parent company Gene by Gene's chief scientist about the SNP's in the chips differing. The SNP lists are of course available on Illumina site, so I guess I can check the lists, but SNPedia probably has some tools with which the SNP-lists for previous platforms were slurper already. I know about Y-SNP, that's why I specifically wrote that FTDNA doesn't provide those SNP's. Allegedly they do also scrub healt-indicated SNP's, which if true, seems pretty inane because every SNP is healt-implicated, eventually by imputation if nothing else. So I guess I'll have to compare the SNP lists now :) --Donwulff (talk) 22:15, 31 May 2014 (UTC)
Statistics for the FTDNA 2014 switcheroo: Human OmniExpress 12 v1.1b had 716,992 probes and Human OmniExpress 24 v1.0a has 713,839 probes. There were 3,153 probes removed; none were added or reported as having different probe sequence. Of the remaining probes, FTDNA currently scrubs all 1,693 Y-DNA probes, 1,148 unknown chromosome (mtDNA?) probes and 2,519 apparently health-implicated probes. Curiously, they instead report 2,479 SNP's which do not have probes on either of these chips. Presumably, they are imputed to fill in holes left by health-SNP scrubbing. If this is indeed the case, it's very ill-advised, because the imputed SNP's will get used to re-impute the health-SNP's or may even accidentally hit actual causal variants, only they will have higher error rate unknown to the user. I would add the SNP lists, but file uploads don't allow straight text or tables, and I'm not sure it makes sense to dump thousands of SNP's into a Wiki page. --Donwulff (talk) 15:41, 2 June 2014 (UTC)
Assumption the SNP's not tested by stock Human OmniExpress are imputed specifically to fill holes left by health-SNP scrubbing (motivated by similar numbers of scrubbed and additional SNP's) seems to fail, by virtue of there being very few if any pairs among them in high LD, and frequently long stretches of additional SNP's between scrubbed ones. Some of the new SNP's don't even exist on any known microarrays, so it's unlikely they're imputed for compatibility with specific chips, or that there even exists microarray probes for those SNP's. They do sometimes seem to occur about 10 kilobases apart; it actually makes more sense for them to be spread evenly out across the genome to smooth out IBD, I was just expecting they would use the information lost from the scrubbing. This would make them even more suspect quality though, as they're just used to fill in blanks, these kinds of imputed data should only be used internally and not released in raw data dumps. I'm unsure how to verify if they're imputed, one possible course is to check from OpenSNP data if they're even in HWE, or have unreasonable LD to surrounding SNP's. Or just ask on FTDNA forums :) --Donwulff (talk) 13:54, 3 June 2014 (UTC)