In the last two years, whole genome association (WGA) studies have identified new candidate genes for several complex diseases, including age-related macular degeneration, diabetes, inflammatory bowel disease, and myocardial infarction. Initial reports have now been published in Molecular Psychiatry for schizophrenia (Lencz et al., 2007) and bipolar disorder (Baum et al., 2007), and several national and international consortia are currently planning or performing very large-scale studies in both disorders. Thus, it is timely to begin considering the potential implications of these investigations.

In their current paper, Baum et al. (2007) present data consistent with a polygenic, common disease/common variant model. No genetic variants of large effect were detected, although the authors duly note that the limitations of the current generation of microarray technology, as well as their pooling approach to genotyping, make a definitive statement premature. For example, the Illumina platform utilized in this study does not cover the pseudoautosomal region that we identified as a possible risk locus for schizophrenia (Lencz et al., 2007).

It is notable that the one SNP (located in the gene encoding diacylglycerol kinase eta, or DGKH) which survived conservative Bonferroni correction demonstrated a modest odds ratio of only 1.59. The risk allele at this SNP, like most of the “top hits” in this study, was very common even in healthy subjects (84.3 percent vs. 89.9 percent for patients). Thus, a sum score combining risk alleles from the top candidates, was fairly high even in controls (though even higher in patients, of course). The authors conclude that their data are consistent with a polygenic threshold model.

Under a polygenic threshold model, the number of possible interactions resulting in a phenotype diagnosed as schizophrenia quickly becomes astronomical. For example, even in an oversimplified model containing only 10 true risk alleles, any four of which could combine to produce illness, the number of possible disease-causing combinations (genetic subtypes) would be 210. Increasing those numbers only slightly, for example, requiring a threshold of eight alleles out of 20 possible risk variants, results in an exponential increase in complexity (>125,000 possible combinations). If possible, analyses that carve out more homogenous subgroups would be helpful. For this reason, it has been suggested that studies with small samples may be a useful complement to large-scale studies, insofar as small studies may (by chance) have improved power at least across a limited subset of the search space, and may avoid potential sources of methodological heterogeneity, such as diagnostic differences across sites (Brzustowicz et al., 2007).

The next challenge will be to characterize these statistical associations at the biological level, in an attempt to identify common biological pathways that may serve as targets for treatment studies. Baum et al. point towards such an approach in their Figure 4, demonstrating that several of their top candidates participate in lithium-sensitive signaling pathways, such as the phosphatidyl inositol signaling pathway. The first major WGA study (Klein et al., 2005) identified a role of the complement system in macular degeneration, leading to invigorated efforts to identify and test related treatments such as complement system antagonists (Kohl, 2006). Hopefully, similar opportunities will soon become available in psychiatric treatment research.

References:

Lencz T, Morgan TV, Athanasiou M, Dain B, Reed CR, Kane JM, Kucherlapati R, Malhotra AK. (2007) Converging evidence for a pseudoautosomal cytokine receptor gene locus in schizophrenia. Molecular Psychiatry, Epub: 03/20/07.

View all comments by Todd Lencz

  I recommend this paper

This paper describes a large, well-conducted, two-stage genetic association study of bipolar disorder using a DNA pooling approach. A genome-wide set of over 500,000 SNPs (Illumina platform) was used in a U.S. discovery sample of 460 cases and 560 controls and 1,877 SNPs were taken forward into a pooling replication experiment in a German sample of 772 cases and 876 controls. The 88 SNPs showing evidence for association in the same direction in both the U.S. and German pools were then typed individually in all the samples. Four SNPs showed combined evidence for association at p This study demonstrates the potential utility of genome-wide association approaches and points to several loci that may influence susceptibility to bipolar disorder. Because of the modest effect sizes of the loci and the oligogenic/polygenic nature of mood and psychotic illness it will be crucial for the reproducibility and generalizability of such findings to be tested across large, adequately powered independent datasets.

As the authors acknowledge, their pooling approach is cost-effective but has reduced power compared with studies that use individual genotyping for all SNPs. Further, individual typing provides much more versatility in terms of analyses that allow more sophisticated approaches to the phenotype, rather than simple case-control analysis.

The imminent availability of multiple genome-wide association datasets for bipolar disorder and schizophrenia will open a new window on the biology underpinning major psychiatric illness. If analyzed optimally, this can be expected to have a major impact on the way we think about and treat mood and psychotic illness.

View all comments by Nick Craddock

  I recommend this paper