Loss-of-function variants in the genomes of healthy humans

Abstract

Genetic variants predicted to seriously disrupt the function of human protein-coding genes-so-called loss-of-function (LOF) variants-have traditionally been viewed in the context of severe Mendelian disease. However, recent large-scale sequencing and genotyping projects have revealed a surprisingly large number of these variants in the genomes of apparently healthy individuals--at least 100 per genome, including more than 30 in a homozygous state--suggesting a previously unappreciated level of variation in functional gene content between humans. These variants are mostly found at low frequency, suggesting that they are enriched for mildly deleterious polymorphisms suppressed by negative natural selection, and thus represent an attractive set of candidate variants for complex disease susceptibility. However, they are also enriched for sequencing and annotation artefacts, so overall present serious challenges for clinical sequencing projects seeking to identify severe disease genes amidst the 'noise' of technical error and benign genetic polymorphism. Systematic, high-quality catalogues of LOF variants present in the genomes of healthy individuals, built from the output of large-scale sequencing studies such as the 1000 Genomes Project, will help to distinguish between benign and disease-causing LOF variants, and will provide valuable resources for clinical genomics.

Figure 1.

Classes of LOF variant affecting protein-coding regions. A model three-exon gene is shown both intact (top) and following the introduction of various types of LOF variant (red triangles). Effects on the transcript produced by the gene are shown at the right. LOF variants typically result in a loss of protein-coding functionality downstream of the variant (red boxes).

Figure 2.

Functional regions are enriched for sequencing errors and other artefacts. Top shows a schematic plot of levels of true sequence variation (blue) and sequencing error (red). Bottom shows pie charts with circle size proportional to total observed variation, and red and blue sections proportional to error and true variation, respectively. In functional regions, true variation is suppressed by natural selection, but error remains approximately uniform. This results in fewer observed variants in functional regions, but a higher error rate in those observed.

Figure 3.

Reported numbers of LOF variants per individual genome in several published large-scale sequencing studies. Individuals labelled European, East Asian or Nigerian are HapMap individuals from reference (27). Numbers for Venter and Lupski are from references (25) and (28), respectively. Small grey segments in Venter and Lupski histograms indicate unreported numbers for splice-disrupting SNPs and frame-shift indels, respectively.