Sunday, July 22, 2007

Reporting genome-wide association studies   posted by p-ter @ 7/22/2007 08:56:00 PM

RPM points to a post from Mark Lieberman at Language Log on the reporting of genome-wide association (GWA) studies. His request (for the popular press; these things are always in the actual paper): report the allele frequency of the associated allele in cases, as well as the frequency in controls. I've often argued that people will get used to the complexity of "complex" disease once they're able to say "Oh, I know several people with the 'diabetes gene' that never got diabetes", but this is a more proactive measure towards that end, and I think it's a great idea.

Lieberman includes the numbers for the "restless leg syndrome"-associated allele I mentioned recently; the other disease I mentioned in that post was gallstone disease, in which the associated allele has a frequency of 10% in cases and 5% in controls. I'll try to remember to report those numbers every time I mention a GWA study from now on.

ADDENDUM: I also like the suggestion from the comments that posterior probabilities be given, as they are sometimes more intuitive. That is, if A is the disease allele and D is disease status, P(A|D) is less interesting, in some sense, than P(D|A). Unforunately, disease prevalances aren't always well-defined, and P(D) is necessary for the calculation. For gallstone disease, P(D) is about 0.15, so P(D|A) = P(A|D)P(D)/[P(A|D)P(D) + P(A|!D)P(!D)] = 26%. The corresponding posterior probability P(!D|A) is 74%. So someone carrying the A allele has a 26% chance of developing the disease, and a 74% chance of staying healthy.

Labels: ,