Monday, April 2, 2012

Ho hum, you're SO close to average!

So a new study confirms (we would claim) what we have been saying for many years,  repeatedly on MT, and in print before it was popular to say it:  Personalized genotyping is not going to be very useful in predicting disease.

The new story in Science shows from twin data--how close can you get!--that overall, individualized prediction of disease is not much different from saying 'you're average'.  Being just like everybody else is not a nice way to feel, but it is what is expected in this kind of situation.  Identical twins are far from identical in terms of their life expectancy of disease history.

Still, appropriate as this story is in refuting the blatantly self-interested claims of many parties in the discussion these days, a Times report of it is also very misleading.  They don't point out the issue.  Neither do the authors of the original paper.

The idea of the twin study is, like heritability studies, that if you look at the whole genome, integrating whatever information it contains, the overall predictive power is slight.  That's essentially because a multitude of minor-effect variants may contribute to risk but individually too little to identify statistically by themselves.  So you look at the aggregate.  But even there, the predictive power is slight.

This is not new!  It's been shown in twin studies before (without having genotype data) and in mouse studies with or without such data.  And it makes sense.

But this does not mean that personalized genotyping is useless.  There are known genes in  which specific mutations confer clearly higher-than-average risk of a given disease (and, should anyone look for them, there must be protective variants as well).

Most twins would receive negative personalized genome tests (that is, reports of risk alleles for what the twin got).  But this doesn't take away from the finding, also long known, that specific high-risk variants with mainly single-gene effects (e.g., cystic fibrosis, some breast cancer variants, etc.) are detectable by genome tests and are useful.

The bottom line:  unless you have such single-gene disorders in your family, generally speaking, skip the test--and the McFood. And if you're worried, go take a walk.

No comments: