Numerous studies have pinned single nucleotide polymorphisms (SNPs) to common diseases such as type 1 diabetes and breast cancer. A new study looks at another major source of genetic variation, copy number variants (CNVs)—extra or missing pieces of the genome. Researchers at the Wellcome Trust Case Control Consortium examined thousands of common CNVs and asked whether they were linked to eight common diseases, from rheumatoid arthritis to type 2 diabetes. They found very few associations, and the three loci they identified had previously been identified in SNP-based studies1. The researchers conclude that “common CNVs that can be typed on existing platforms are unlikely to contribute greatly to the genetic basis of common human diseases.” We asked three experts about the implications of the results and whether they agree with this conclusion.

Gonçalo Abecasis:

It has long been known that rare structural variants can contribute to human disease, and recent examples—such as the association of variants in the neuronal growth regulator-1 (NEGR1) locus with obesity and the LCE gene cluster with psoriasis—had suggested that common structural variants might be major contributors to complex disease susceptibility. This is the most thorough and systematic assessment of the contribution of common structural variants to complex disease, and it shows they are unlikely to be major contributors to disease susceptibility. In some cases, such as the examples the authors point to in the human leukocyte antigen (HLA) region, the structural variant association signals might be shadows of much stronger nearby signals resulting from SNPs or other types of variants.

Importantly, this work also illustrates that current methods for assessing structural variation are very sensitive to artifacts due to sample handling and processing. The potential for these artifacts emphasizes that, just like other large association studies, structural variant association should employ a full array of quality-control protocols for results—ranging from Q-Q plots to careful evaluations of missing data rates by sex or sample source. Unless it can be shown that cases and controls were handled exactly in parallel, and that potential artifacts have been excluded, the results of any future structural variant association study must be interpreted with a large grain of salt.

Felix Moore Collegiate Professor of Biostatistics, Center for Statistical Genetics, University of Michigan School of Public Health, Ann Arbor, Michigan, USA.

Joseph Nadeau:

A major problem in modern genetics is explaining 'missing heritability'—the striking and unexpected difference between the often strong heritability of a trait and the consistently modest cumulative variance explained by genetic variants (SNPs) detected in genome-wide association studies. The expectation had been that common SNPs that usually have weak phenotypic effects would together account for phenotypic variation and susceptibility to common diseases. But the discovery of 'missing heritability' indicates that this model is inadequate.

An alternative model invokes other kinds of genetic variants, such as CNVs. But, disconcertingly, these, too, are exonerated, as this study nicely demonstrates. Related studies are pursuing other hypotheses by reevaluating heritability estimates, exploring unexamined regions of the genome, analyzing other classes of genetic variation and testing other genetic models based on less common (rare) variants that have strong phenotypic effects. But I have a hunch that, together, these studies will still not fully account for 'missing heritability'. Perhaps it is time to consider less conventional explanations such as heritable epigenetic changes, a possibility bolstered by exciting new evidence in many model organisms.

Chair, Department of Genetics, Case Western Reserve University School of Medicine, Cleveland, Ohio, USA.

David Goldstein:

Common CNVs account for much of the (per base) heterozygosity in people, are often definitively functional and are now shown to be largely irrelevant in many common diseases. The limited impact of common CNVs is particularly striking given the explosion of associations between rare CNVs and many neuropsychiatric diseases.

Part of the reason for the largely negative results could be the mix of diseases under study, but the key reason is surely the focus on common CNVs. It seems fair to conclude that, for the most part, disease-causing CNVs are under strong negative selection and therefore cannot rise to even moderate frequency in the population. The contribution of CNVs to disease risk therefore seems to reflect a classic mutation-selection balance, with mutation introducing harmful CNVs and selection eliminating them.

Although these conclusions are not unexpected, given that common CNVs are known to be tagged by common SNPs covered by genome-wide association studies, it is nonetheless useful to have a direct and comprehensive analysis of common CNVs. Still, one can't help but wonder, what might have been found if the focus had been on precisely the opposite class of CNVs, those with a frequency of less than 5%? Many of the eight diseases considered here have been much less studied for rare large (in terms of base pairs) CNVs than diseases such as autism and schizophrenia have. For this reason, we remain ignorant of whether the diseases studied by the Wellcome Consortium differ in the degree to which they are influenced by CNVs or whether the apparent increased importance of CNVs in neuropsychiatric disease simply reflects ascertainment differences.

Director, Center for Human Genome Variation, Duke University, Durham, North Carolina, USA.