Puzzling the odds of disease risk

By Razib Khan | March 23, 2011 12:42 pm

I don’t currently put much stock in disease risk estimates given by personal genomics firms. My pedigree is pretty large, so I already have a lot of information on the table (though the demographic transition across the generations in Bangladesh is pretty evident in my own family). Family information is currently much more useful in estimating relative risks than GWAS for most diseases (as it is for quantitative traits like height and I.Q.) If a friend wants to get genotyped to better understand their disease risks, unless they’re rolling in money, I generally advise them to hold off.

But the returns on personal genomics in this area has nowhere to go but up, so we should understand exactly what the odds for disease risk we are getting from firms like 23andMe now, because we will have to at some point in the future. Carl Anderson at Genomes Unzipped has the first post of two up on this issue. The plot I’ve taken from the post shows the dependence of relative risk as a function of number of risk loci included. Remember that the average risk for the trait, ulcerative colitis, is rather low, at ~1%. So even if you have a 2.5 × greater than average risk…I wouldn’t stress out too much.

Note that the relatively marginal predictive value of personal genomics currently means that worries about this sort of thing making people uninsurable are probably not grounded in reality. The family disease history is much more actionable when estimating your premium, or whether they should underwrite you in the first place.

CATEGORIZED UNDER: Genetics, Genomics, Personal Genomics

Comments (7)

  1. Zachary Kurtz

    One thing that I’ve noticed from my brief time with my 23andme results, is that people are [correctly] more concerned when they get back a high(er) probability of disease risk, and that this concern leads them to become overly or underly confident that the measurement is accurate (by either freaking out or ignoring it).

    The accuracy of the test, largely, shouldn’t depend on the actual the result, but that’s just seems to be how people take extreme news.

    Any data on this?

  2. mem

    Can you really assure me that insurers won’t misuse the information anyway? Promise?

    Sure, they shouldn’t. But they do a lot of things are are not nice.

  3. first, as i asserted above, right now they can’t do much with it cuz there isn’t much they don’t have already. second, “misuse” is a loaded term. health insurance is a bizarre industry, and in the USA it isn’t viewed as real insurance anyhow, with all the regs there are.

  4. mem

    Well, I think “not grounded in reality” is kind of a loaded term too. I’ll bet you haven’t had to try to get insurance on the US open market. From someone grounded in that reality: they aren’t trustworthy.

  5. you just lost that bet.

  6. DK

    What would be interesting (and very simple) to see is this:

    Take DNA of all the dead people with good medical history and see if aggregate of all the various risks predicts time and/or cause of death with any reliability. $200K buys N=2,000 which is not unreasonable for entities that have access to lots of dead peoples’ DNA.


Discover's Newsletter

Sign up to get the latest science news delivered weekly right to your inbox!

Gene Expression

This blog is about evolution, genetics, genomics and their interstices. Please beware that comments are aggressively moderated. Uncivil or churlish comments will likely get you banned immediately, so make any contribution count!

About Razib Khan

I have degrees in biology and biochemistry, a passion for genetics, history, and philosophy, and shrimp is my favorite food. In relation to nationality I'm a American Northwesterner, in politics I'm a reactionary, and as for religion I have none (I'm an atheist). If you want to know more, see the links at http://www.razib.com


See More


RSS Razib’s Pinboard

Edifying books

Collapse bottom bar