An analysis of results from 23andme | Hacker News Reader

An analysis of results from 23andme

nsaunders.wordpress.com

38 pointsbbgm9 comments

The use of GWAS data to infer global ancestry is quite entertaining, and I'm glad that the author shared so much about 23andme's results with us.

Still, I have the urge to mention this everytime I see 23andme, so apologies in advance. Your genetic profile is currently less useful than a proper history and physical exam (but mostly just history) in predicting risk for the vast majority of diseases.

This may change, but the extent to which the GWAS data from 23andme will be useful for imputation of the (as yet undiscovered) causal variants that we'll be discovering in the near future is currently unclear.

So in the meantime, enjoy every aspect of their services, but I would encourage you to put little emphasis on their disease predictions. The fact that you don't have one of the known breast cancer mutations doesn't mean that you don't have a novel one. The fact that you are predicted to have high risk of heart disease doesn't mean that you don't have a novel cardioprotective allele elsewhere. At some point in the near future, genetic data will be really useful for risk prediction. We're not quite there yet with the GWAS data that 23andme uses (EDIT or any GWAS data - this is no knock on 23andme).

The reason I posted this particular one is that the author is well aware of the limitations. The reason it's interesting is that 23andme and their ilk do make a certain level of information more accessible, which is not a bad thing, and hopefully makes people more curious.

On GWAS, while the utility of GWAS alone is questionable, GWAS is and will be a key part of many discovery programs for a while. It's tractable, and combined with expression does provide some fascinating insights into disease.

Yes, thanks for being selective! I appreciated the author's relatively minor discussion of the disease data and more expansive discussion of the rest. For completeness: when I say 'GWAS' in this context I am referring to 'common SNPs' (MAF>=5%, or perhaps down to 1%), not to genome-wide association studies based on the SNPs. I used to argue with my labmates about the proper use of the term 'GWAS': I felt that it should refer to the studies only, and 'common SNPs' should be called 'common SNPs'; my labmates preferred to call common SNPs 'GWAS data' and I've finally been converted.

Anyways, I do think that expression+GWAS is useful, but it's not going to get us to a place that's better than history for most diseases.[1] Exome and then genome sequencing will be needed for that.

[1] = On the other hand, Purcell has an interesting paper out that discards the genome-wide significance threshold, with interesting results.

No disagreement on any points :)

You should also use Promethease which automatically links your raw data to the snpedia data to product a really long report of various traits and research results.

http://www.snpedia.com/index.php/Promethease

It's a bit messy and hard to interpret but you get a lot more the info that 23andme provides.

my take on 23andme: http://paulstamatiou.com/review-23andme-dna-testing-for-heal...

I run a DNA test in 2005. I'm chilean and have british ancestors and I wanted to find out what was my origin (English, Scottish, Irish). A Carpenter 'cousin' from USA told me about DNA for Genealogy and bought me a DNA kit from Family Tree DNA http://www.familytreedna.com

I didn't match any Carpenter from USA but a lot of people from UK and USA. Genealogy is a fascinating hobby.

I got my results about 20 minutes ago. (I should have got it a few weeks ago, but DC post office screwed up delivery of the spit kit.) I assume a lot of other people on HN have gotten results recently: http://news.ycombinator.com/item?id=1288125

Very interesting analysis.

On a sidenote, are there any alternatives to 23andme that are as good?

Reply on news.ycombinator.com