Epistasis Blog

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

Tuesday, April 03, 2012

The predictive capacity of personal genome sequencing and missing heritability

There are two papers I would like to call your attention two that provide fuel for a healthy debate about the complexity of the genotype to phenotype mapping relationship. The first, from Eric Lander's group (Zuk et al. 2012), calls into question how heritability is estimated. This is an interesting paper that makes the argument that epistasis or gene-gene interaction is likely to explain a significant amount of the missing heritability. Many heritability estimates are based only on additive effects. My only concern with this paper is that this is not a new observation. My group and many others have been writing about the importance of epistasis for many years and much of this work is not acknowledged. The second paper, from Roberts (2012), calls into question the usefulness of personal genomic sequencing for preditcing disease. This paper shows that it is difficult to predict disease in MZ twins. In other words, twins do not always die of the same diseases. These papers both call into question the assumption made by genome-wide association studies (GWAS) of common alleles and whole-genome sequencing for rare alleles that there will be single loci with big effects that will in turn be useful for personalized medicine.

Zuk O, Hechter E, Sunyaev SR, Lander ES. The mystery of missing heritability: Genetic interactions create phantom heritability. Proc Natl Acad Sci U S A. 2012 Jan 24;109(4):1193-8. [PubMed]

Roberts et al., The predictive capacity of personal genome sequencing. Science Transl Med (2012), in press. [PubMed]

Here are several other blog posts and new stories that provide positive and negative perspectives on the Roberts/Vogelstein paper. It is interesting to note that a number of geneticists are upset that the paper got published and that it received so much press (e.g. Nature News Blog). I think they are more worried about the negative publicity for genome technology than the actual science (which they claim is flawed).

Alzheimer Research Forum

Genomes Unzipped

Nature News Blog

Science News


At 11:11 PM, Blogger Unknown said...

Hi. I'm a graduate student at UCI and read your blog regularly. It's a great resource. I just read the Roberts et al. paper and believe there is a significant flaw. They have assumed that the "genometypes" -- essentially clusters of genomes from monozygotic twins that share a similar genetic risk for a disease -- are completely independent from other genometypes. That is, they ignore, for example, the fact that a genometype with 70% risk may share a signification portion of the internal structure (combinations of SNPs, for example) as one with a 20% risk. Thus, genometypes should not be truly considered (like they have done) independent entities. The problem in their model arises when this faulty assumption of independence is passed on to disease probabilities in which they take the sum over all genometypes individually (equations 6a-8, pages 5-6). This has to affect their results, especially that their claim that epistasis is completely taken into account when estimating the maximum possible reliability of a whole genome sequencing test. This paper makes me believe (like you've been saying all along) that more information (i.e., genetic architecture) is needed in conjunction with whole genome sequencing in order to predict disease risk for a population.


Post a Comment

<< Home