Update to Terwilliger and Göring's "Gene Mapping in the 20th and 21st Centuries" (2000): Gene Mapping When Rare Variants Are Common and Common Variants Are Rare

Our Lasker Prize winning article from 2000 was a discussion of the myriad factors that influence the success or failure of gene mapping studies, and in it we suggested that the reasons we geneticists have had so little success in unraveling the etiology of complex traits had more to do with biology than technology and that many of the decisions about future strategies were being made based on unrealistic and untested assumptions. At the time, the common variant/common disease (CVCD) hypothesis (Reich and Lander 2001) was widely believed to provide a plausible model for complex disease genetics, in which common alleles [defined at the time as greater than 0.10 MAF (minor allele frequency)] with high attributable fractions, such as APOE4 and Alzheimer disease, would dominate trait etiology (Fallin et al. 2001; Pennisi 1998). As argued by Risch and Merikangas (1996: 1516), "Despite the small effects of such genes, the magnitude of their attributable risk may be large because they are quite frequent in the population, making them of public health significance." The discussions at the time proposed that such alleles should be ubiquitous across world populations, predating the expansion of human populations out of Africa (Goldstein and Weale 2001; Lander 1996). We argued that this was inconsistent with what we know about the nature of the evolution of genetic variation (Terwilliger 2001; Terwilliger and Weiss 1998; Weiss and Terwilliger 2000) and "that a larger number of less common alleles is likely to be involved in the etiology of complex disease" (Terwilliger and Göring 2000: 122) with individually trivial attributable fractions. At the time, we were accused of being naysayers and extremists for suggesting that genome-wide association studies would not be uncovering APOE-like effects for schizophrenia and most other complex traits (Brookes 1999; Emahazion et al. 2001; Pennisi 1998). Leading proponents of the CVCD hypothesis predicted in 1998 that "it would be surprising if most of the major genetic factors involved in human disease were not defined over the next 5–10 years" (Bell 1998: 618) and that "the widespread redefinition of disease through genetics will be accompanied by the use of genetics for prediction and diagnosis and to optimize treatment in most common diseases. This is likely to occur within the next decade" (Bell 1998: 620).

Ten years and billions of taxpayer dollars later, our once "extreme" position has replaced the mainstream opinion of a decade ago, as the recent flood of empirical data from genome-wide association studies consistently yields results that are inconsistent with the CVCD hypothesis and rosy scenarios about personalized medicine. In recent months, papers have been trumpeting the success of genome-wide association studies in finding new rare (i.e., p < 0.1) variants with individually minuscule attributable risks, forcing investigators to admit that their results effectively prove that there must be hundreds or thousands of alleles of small effect influencing schizophrenia and bipolar disorder (Purcell et al. 2009), height (Aulchenko et al. 2009), hypertension (Levy et al. 2009), type 2 diabetes (Zeggini et al. 2007), and virtually every other complex trait examined. Perhaps the only common variant of high attributable fraction that has proven to be a useful predictor of disease in the past decade is the lactase-persistence allele identified through family studies in Finland (Enattah et al. 2002), and the allelic simplicity of this locus has more to do with the fact that the new mutation added a useful function—ability to digest lactose in adulthood—rather than making a gene work less efficiently, as most random mutations tend to do. Even the most trumpeted "success stories" in genome-wide association studies, such as Crohn's disease (Barrett et al. 2008), end up implicating tens of specific rare variants, which put together explain only a few percent of the overall genetic contribution to the disease, putting the promised personalized medicine and "gene-chip...