4 ms·
> unfortunately I can't tell from the article or the journal abstract how big the SNP sets involved actually are, but given the difficulty in finding them I'd e
by gwern 12y ago
> unfortunately I can't tell from the article or the journal abstract how big the SNP sets involved actually are, but given the difficulty in finding them I'd expect more than four or five SNPs per set.
Table 1 ( https://pdf.yt/d/rErZm8HUluBbj8xu https://pdf.yt/d/rErZm8HUluBbj8xu / https://dl.dropboxusercontent.com/u/5317066/2014-arnedo.pdf https://dl.dropboxusercontent.com/u/5317066/2014-arnedo.pdf / https://www.sendspace.com/file/ht7shm https://www.sendspace.com/file/ht7shm ) list characteristics of the sets; they range from a low of 3 to a high of 32, with associated risks of 70 - 100% (!).
The size of the sets with a risk of schizophrenia between 90% and 100% is 11/11/19/4/6/3/10/6/9. (SNP set 87\_76 is the one with just 3 SNPs and a risk of 95%. Very nasty.)
Unfortunately I don't think this risk information is enough to to calculate how much predictive value SNPs would have for someone undiagnosed since you'd need to know how many healthy people have these SNP sets as well? But I'm sure the genomics hobbyists and anyone with 23andme data are probably looking closely at the paper and trying to extract risk estimates...