Epistasis Blog

From the Computational Genetics Laboratory at the University of Pennsylvania (www.epistasis.org)

Tuesday, January 19, 2010

Genetics of diabetes reveals biology but does not improve prediction

I very much enjoyed this blog posting on www.phgfoundation.org. They discuss a new paper published in the British Medical Journal (below) that shows traditional risk factors do a much better job of predicting Type II Diabetes than 20 published SNPs. A quote from the post: "By assessing the area under the receiver operator characteristic curve (a plot of sensitivity versus 1-specificity, where a value of 1.0 represents a perfect test and 0.5 represents a useless test), the traditional models significantly outperformed the genetic model (around 0.75 versus 0.54), and their performance was not substantially improved by the addition of genetic risk factors." This comes as no surpise to me because the genetic studies that led to this test were all based on single-locus analyses that completely ignore the underlying complexity of this common disease. It is my working hypothesis that we will not be able to use genetic to predict disease risk until we ebrace, rather than ignore, the complexity of the genetic architecture of common human diseases. We commented on this in a 2007 letter to Science (also below).

Talmud PJ, Hingorani AD, Cooper JA, Marmot MG, Brunner EJ, Kumari M, Kivimäki M, Humphries SE. Utility of genetic and non-genetic risk factors in prediction of type 2 diabetes: Whitehall II prospective cohort study. BMJ. 2010 Jan 14;340:b4838. doi: 10.1136/bmj.b4838. [PubMed] PMID: 20075150.

Williams SM, Canter JA, Crawford DC, Moore JH, Ritchie MD, Haines JL. Problems with genome-wide association studies. Science. 2007 Jun 29;316(5833):1840-2. [PubMed] PMID: 17605173.

0 Comments:

Post a Comment

<< Home