3 ms·

Ah, you are correct. 23andMe is using Illumina, Navigenics is using Affy. I have my doubts as to whether either approach is truly useful at this point in time

by apathy 19y ago

Ah, you are correct. 23andMe is using Illumina, Navigenics is using Affy. I have my doubts as to whether either approach is truly useful at this point in time to non-researchers, but I could be quite wrong. My objection to 23andMe and Navigenics and so forth is based on the inherently incomplete information provided by a static snapshot of a person's polymorphisms. As far as email -- sure. Though I don't know whether I'll be pleased to be associated with such a screed when the full details of their plans are more apparent. Selling their database to pharma companies is exactly the sort of evil I'd hoped they'd avoid. Oh well. Looks like the Google way of thinking didn't infect 23andMe after all.

rms 19y ago
I can't find the link on their site now, maybe it's gone, but I also believe that 23andme is using a modified Illumina chip that is more inaccurate than the standard chip but is 10 times more likely to return SOME result. Here it is: https://www.23andme.com/you/faqwin/dataaccuracy/ https://www.23andme.com/you/faqwin/dataaccuracy/
- apathy 19y ago
  Not sure that it matters -- if you have a less accurate chip, you just resequence (run it) to greater depth. (If you can afford to, that is... and in all likelihood, they can) That's effectively what Helicos et al. are doing anyways. Their call rate for whole-genome resequencing is on the order of 99.7% at a depth of 2; at a depth of 20-50 maybe it becomes useful. If you think of the chip (hybridization) methods being depth of 1, eg. a single scan per sample across however many probes there are for a given SNP, then you can probably see where the synthesis methods win. It's really cheap to re-run them. It's twice as expensive to run 2 chips... All of the sequencing-by-synthesis methods are less accurate but cheaper and ultimately more useful than anything based on the existing hybridization arrays (or beadarrays). If you look at eg. the Personal Genome Project, there is a good summary from the Church lab about how eg. padlock sequencing is different from Affy/Illumina/Nimblegene hybridization arrays. Also note that (1 - 0.999) * 550K still gives 550 miscalls. D'oh. (Another big win for sequencing by synthesis -- just do it again, and again, and again...) This is what I'm getting at when I bitch about 23andMe not advancing the state of the art. With the money they have, they could. But they're not.

by apathy 19y ago

rms 19y ago

I can't find the link on their site now, maybe it's gone, but I also believe that 23andme is using a modified Illumina chip that is more inaccurate than the standard chip but is 10 times more likely to return SOME result. Here it is: https://www.23andme.com/you/faqwin/dataaccuracy/ https://www.23andme.com/you/faqwin/dataaccuracy/

apathy 19y ago
Not sure that it matters -- if you have a less accurate chip, you just resequence (run it) to greater depth. (If you can afford to, that is... and in all likelihood, they can) That's effectively what Helicos et al. are doing anyways. Their call rate for whole-genome resequencing is on the order of 99.7% at a depth of 2; at a depth of 20-50 maybe it becomes useful. If you think of the chip (hybridization) methods being depth of 1, eg. a single scan per sample across however many probes there are for a given SNP, then you can probably see where the synthesis methods win. It's really cheap to re-run them. It's twice as expensive to run 2 chips... All of the sequencing-by-synthesis methods are less accurate but cheaper and ultimately more useful than anything based on the existing hybridization arrays (or beadarrays). If you look at eg. the Personal Genome Project, there is a good summary from the Church lab about how eg. padlock sequencing is different from Affy/Illumina/Nimblegene hybridization arrays. Also note that (1 - 0.999) * 550K still gives 550 miscalls. D'oh. (Another big win for sequencing by synthesis -- just do it again, and again, and again...) This is what I'm getting at when I bitch about 23andMe not advancing the state of the art. With the money they have, they could. But they're not.