Comments on: We need to create a market for genetic-association data

By: EntityTib

EntityTib — Wed, 20 May 2009 16:27:11 +0000

Good Day:
I enjoy reading your blog. Thanks.

By: Lee Watkins

Lee Watkins — Mon, 16 Feb 2009 19:35:48 +0000

An author’s lay-oriented summary would be a step forward (since it’s almost never done now) and could be very helpful. But it would only refer to the current state of affairs at that point in time and would shortly become obsolete and/or misleading given how quickly things change. To be more useful over time these summaries would either have to be periodically updated in light of new data, or be somehow dynamic and/or self-maintaining (something for which I can’t quite envision a practical mechanism, semantic/syndicated or otherwise).

That’s kind of the problem now with using any genetic-association collection (SNPedia, NIH’s geneticassociationdb, etc.) – they’re only as useful as their most recent data and how well it’s integrated with the larger body of genetic-association evidence. Separating the good/accurate from the bad/inaccurate information may not be easy, nor is understanding the implications of that information.

Even current, accurate genetic-association data may have limited usefulness given how little predictive value they usually seem to have (as jor’s comment alludes to), with few exceptions where effects are strong and straightforward, at least until lots more significantly-associated polymorphisms are discovered for a given disease or trait. Even then their effects would probably have to be considered in aggregate / cumulatively rather than individually. Interesting that NIGMS just funded InSilicos to work on software for roughly this sort of thing using a statistical technique to “produce stable predictions from data with a large number of dimensions … as a way of using these signals to predict disease.”

By: jor

jor — Tue, 03 Feb 2009 04:37:50 +0000

I would take a completely different view point — sequence information so far has been relatively clinically useless and to some extent disappointing (look at odds ratios published, they are tiny). Maybe the magic of the genome is around the corner — but for now this sounds like an expensive way to provide jobs for bioinformaticians.

The information is becoming a commodity, but so is whole body CT scan — and we don’t hear many proponents for that these days.

By: Lynn Dowling

Lynn Dowling — Mon, 01 Dec 2008 19:40:33 +0000

Bravo Doctor Altman: I am working with a community hospital to develop a clearinghouse of genetic tests considered “ready for prime time” for our medical staff and patients to access, with assistance from genetic counselors. Everyone overlooks hospitals as a source for disseminating such information, yet every hospital in the country “aggregates” hundreds of physicians and serves as a source of health information for their communites. Perhaps your idea would be more readily adopted if one of the big hospital associations would make the dissemination of such summaries available to its members. We would surely subscribe!

By: dr s.m.

dr s.m. — Wed, 19 Nov 2008 16:06:12 +0000

Nigam: Nature news says “Smart Genetics chief executive Julian Awad found himself in a controversy over whether it violated intellectual-property agreements covering APOE testing.”

Assuming Alzheimer Mirror tests APOE e4, nobody in the field doubts that association. Perhaps you have secret information?

http://www.nature.com/news/2008/081029/full/4551155a.html

By: Nigam Shah

Nigam Shah — Fri, 14 Nov 2008 00:45:54 +0000

It might also make sense to call for a “commons” for such data (in the sense that Lawrence Lessig puts it). Right now, each company such as 23andme, navigenics, smart genetics and the 10s of wannabes of genetic testing are “curating” the same association and interpreting it in widely different ways. (e.g. the Alzheimer’s mirror test by smart genetic just had to be pulled from the market because of unreliable association data).

It is a shame that no information models exist for communicating such associations and everyone has their own ad-hoc interpretation mechanisms.

Having a “commons” where the authors write the interpretation in a syndicatable format solves both these issues. Then everyone is free to dip into the commons and offer their analysis or interpretive services on it. Those that don’t need help interpreting don’t need to pay for it.

By: Pierre Lindenbaum

Pierre Lindenbaum — Thu, 13 Nov 2008 19:59:05 +0000

As a bioinformatician, I am also dreaming of semantic abstracts (http://tinyurl.com/5f9p9k). The RDFa specification could be of help here but I don’t think that the NCBI is ready to make some changes. I’m also working with association studies and here I would dream of an RDF/XML format to share the informations about the individuals, the markers and the genotypes instead of those ugly ‘linkage’ files filled with a bunch of ‘0′, ‘1′ and ‘2′….

By: Steven Murphy M.D.

Steven Murphy M.D. — Thu, 13 Nov 2008 19:23:30 +0000

Dr. Altman,
Fantastic Idea, we are trying to do just that at HelixGene……The more participants we have…the better….True, there is EGAPP, but this is more than that…

-Steve
http://www.thegenesherpa.blogspot.com