The publication of this first genome-wide single nucleotide polymorphism (SNP) association study in Parkinson disease (PD) has created considerable debate in the field. The most public parts of this discussion include four follow-up articles (Clarimon et al., 2006; Farrer et al., 2006; Li et al., 2006; and Gorris et al., 2006) and a short introductory piece (Myers, 2006), all currently in the early view of the American Journal of Human Genetics. (I should make it clear that I am a coauthor on one of these studies, Clarimon et al., which is largely a failure-to-replicate piece).
The discussion centers on two aspects of the original and subsequent studies. First, that Maraganore and colleagues’ study was questionable because follow-ups failed to identify association in positive SNPs. Second, that because this represents such a key resource, the raw genotype data must be released, in its entirety, into the public domain.
Dealing with the first issue, describing the Maraganore et al. study as a failure because of lack of replication is an oversimplification of this type of experiment. There is a need to understand what this approach can and cannot tell us. For instance, genotype association will not reveal relatively rare mutations that lead to disease (such as those in synuclein, parkin, or even LRRK2 in North Americans). Instead, the design of these experiments is to identify common genetic variability underlying disease. By its very nature, high-density genome-wide SNP association results in many false positives. Where the genetic variability underlying disease is only of minor to moderate effect (as is suspected in PD), the level of "noise" and true signals will be about equivalent, and the former will outweigh the latter. The most conservative method to sort through this noise is multiple rounds of replication in independent cohorts. Thus, it is to be expected in this scenario (and in the absence of an overwhelming positive signal) that the vast majority of initial hits will be false positives. Identifying true positives will take a large-scale screening of numerous significant loci, in several series.
The absence of a single overwhelming positive in the initial experiments may at first blush be disappointing, but this, too, misses the beauty of this relatively unbiased type of experiment. With the caveats of genomic coverage (i.e., how much of the genome was successfully assayed by Maraganore and colleagues' approach), these data tell us not only what might be there, but also what is not there. It suggests that in PD there is not a single common variant of large effect, such as ApoE in Alzheimer disease. This is invaluable data when we come to consider and design experiments.
As for the second issue—release of data to the public domain—large-scale genetics such as that described by Maraganore et al. require considerable resources, not only to perform the experiments ($100,000 to million-dollar experiments), but also to manipulate and analyze the results. Given the investment, there can be little doubt that funding agencies will make it a requirement that raw genotype data (not just p values) are made publicly available at publication. All internal review board and intellectual property issues relating to such data should be ironed out prior to funding. Maraganore and colleagues have stated that they will be releasing their data publicly. There are currently resources that support posting of these data (http://www.ncbi.nlm.nih.gov/WGA/), or that are already posting this type of data (http://ccr.coriell.org/ninds/), so one would hope public release will be very soon.
Jordi Clarimon, Sonja Scholz, Hon-Chung Fung, John Hardy, Johanna Eerola, Olli Hellström, Yi-Ru Wu, Pentti J. Tienari, Andrew Singleton. Conflicting results regarding the Semaphorin gene (SEMA5A) and the risk for Parkinson's Disease
Matthew J. Farrer, Kristoffer Haugarvoll, Owen A. Ross, Jeremy T. Stone, Nicole M. Milkovic, Stephanie A. Cobb, Andrew J. Whittle, Sarah J. Lincoln, Mary M. Hulihan, Michael G. Heckman, Linda R. White, Jan O. Aasly, J. Mark Gibson, David Gosal, Timothy Lynch, Zbigniew K. Wszolek, Ryan J. Uitti and Mathias Toft. Genome-wide association, Parkinson's disease and PARK10
Yonghong Li, Charles Rowland, Steven Schrodi, Walter Laird, Kristina Tacey, David Ross, Diane Leong, Joseph Catanese, John Sninsky, Andrew Grupe. A case-control association study of the twelve single nucleotide polymorphisms implicated in Parkinson’s disease by a recent genome scan
A. Goris, C.H. Williams-Gray, T. Foltynie, D.A.S. Compston, R.A. Barker, S.J. Sawcer. No evidence for association with Parkinson disease for 13 SNP's identified by whole genome association screening
References:
Jordi Clarimon, Sonja Scholz, Hon-Chung Fung, John Hardy, Johanna Eerola, Olli Hellström, Yi-Ru Wu, Pentti J. Tienari, Andrew Singleton. Conflicting results regarding the Semaphorin gene (SEMA5A) and the risk for Parkinson's disease.
Matthew J. Farrer, Kristoffer Haugarvoll, Owen A. Ross, Jeremy T. Stone, Nicole M. Milkovic, Stephanie A. Cobb, Andrew J. Whittle, Sarah J. Lincoln, Mary M. Hulihan, Michael G. Heckman, Linda R. White, Jan O. Aasly, J. Mark Gibson, David Gosal, Timothy Lynch, Zbigniew K. Wszolek, Ryan J. Uitti and Mathias Toft. Genome-wide association, Parkinson's disease and PARK10.
Yonghong Li, Charles Rowland, Steven Schrodi, Walter Laird, Kristina Tacey, David Ross, Diane Leong, Joseph Catanese, John Sninsky, Andrew Grupe. A case-control association study of the twelve single nucleotide polymorphisms implicated in Parkinson’s disease by a recent genome scan.
A. Goris, C.H. Williams-Gray, T. Foltynie, D.A.S. Compston, R.A. Barker, S.J. Sawcer. No evidence for association with Parkinson disease for 13 SNP's identified by whole genome association screening.
View all comments by Andrew Singleton
Home
