|
|
|
|
|
|
by maxander
3883 days ago
|
|
|
> "We are supposed to believe that the technology can tell us accurately if a person has 1 copy of a variant, but it cannot determine if there are 2 copies? What does 23andMe report if there are 2 variants? Nothing?" Actually, this sounds eminently plausible. As I understand it, a DNA microarray consists of a bunch of probes that each transmit a signal when they bind to a specific strand of digested DNA. A probe is either bound or not, and since you're applying a mixture obtained from multiple cells, there's many copies of each DNA sequence present in either case. There's no way you could establish whether or not two copies of the same gene were present in an individual cell. I could be wrong (I'm a bioinformatician, not a biologist), and 23andme could be doing something interesting which reveals this data- in either case, I'd be interested to hear. But this makes fairly bad advertising copy- everyone else with the same cursory understanding of this technology is going to have the same thought, and think that you don't even have said cursory understanding if you don't explain your reasoning better [1]. [1] For that matter, you come off as wildly unscientific simply because you make this sort of conjecture without citing sources. There is a resource out there, somewhere, that would confirm or deny what you're saying, and your ability to make a credible scientific statement rests on you finding it. |
|
|
There is a nice animation of the technology on Illumina's website. It's a single base extension assay where you have hybridization of a probe, then add a labeled nucleotide where the SNP is that you are studying. For a A/T SNP, if you basepair and add an 'A' you get a green signal, if you add a 'T' you get a red signal. If you end up with a mix of green/red, you have a heterozygous variant.
Illumina has a tech note on the call rate and error rate here: http://www.illumina.com/Documents/products/technotes/technot... At a minor allele frequency of 0.5% their minor allele homozygous error rate was 0.1%.
The nice thing about this technology is that the more data you have, the better your calls get, and 23andMe has real world data on over 1 million customers. (Illumina tech note used data from 2,000 genotyped samples) I think 23andMe has enough data to know which of the SNPs on their genotyping chip are accurate, but they don't exactly open up their data for public inspection.
If you are a 23andMe customer you can see the improvements to the raw data they have made over the years here: ( https://www.23andme.com/you/download/revisions/ ) If you can't see it, it says things like: "July 28th, 2014. Analysis of our data has allowed us to improve the interpretation of over 10,000 SNPs genome-wide on the V4 chip." When they get more data, they are improving their calls.
Finally, the FDA document about 23andMe's approved Bloom Syndrome carrier test says that "all homozygous variant genotype samples receive a 'no-call' result, since the calling software was designed not to detect homozygous variant genotypes." It sounds to me like they designed the software to ignore and throw out homozygous data. http://www.accessdata.fda.gov/cdrh_docs/reviews/DEN140044.pd...