Introduction
Studies using deep and population-scale sequencing have reported large numbers of rare variants (minor allele frequency [MAF] < 1%) present as a consequence of recent and rapid human population expansion.1–6 However, interpreting the impact of rare variation remains an ongoing challenge. Several exome sequencing studies have suggested that rare variants are of broad importance with the finding that they represent the majority of potentially deleterious and damaging protein-coding alleles2 and can contribute to complex disease risk.7–11 In contrast, population-genetic models have indicated that rare alleles are unlikely to be large overall contributors to heritable variation for many complex diseases.12 Indeed, large population studies of rare variants in autoimmune disorders have so far found negligible impact,13 and analyses of personal genomes have reported multiple rare and protein-code-disrupting sites in presumably healthy individuals.14,15 Further compounding the challenge of understanding the impact of rare variation has been that most studies have focused on only protein-coding alleles whose interpretation is facilitated by the genetic code. For rare variants in noncoding regions, there is no analogous code to aid in the prediction of their impact even though these regions harbor considerable complex-disease-associated variation16,17 and most likely contain an abundance of important rare alleles.
Currently, genetic studies of gene expression provide a systematic means of identifying functional noncoding variants; such studies have identified noncoding variants associated with gene expression, splicing, and allele-specific expression (ASE).18–20 However, insight into the impact of rare noncoding variants has been limited. Few studies have had the advantage of full genome sequencing data and, even when these data are available, they have only assayed unrelated individuals, providing limited power to describe rare-variant effects.18,21,22 To overcome this challenge and provide more systematic insight into the impact of rare noncoding variants, we coupled high-quality genomes with transcriptomes within a large family (n = 17 individuals). The advantage of this design is that the large number of children (n = 11) provides high-confidence rare variants established through both deep sequencing and Mendelian segregation as well as sufficient power to test for cis-expression quantitative trait loci (eQTLs) present within a single human family. Furthermore, eQTLs from the family can be compared to eQTLs from a cell-type- and ethnicity-matched population sample recently reported by the Geuvadis Consortium,18 providing the unique ability to identify large genetic effects specific to the family and test their relationship to rare variants.23 Indeed, we report that rare regulatory variants are enriched near genes that exhibit large-effect cis-eQTLs for gene expression, splicing, and ASE within the family. Furthermore, the family eQTL genes are more evolutionarily constrained than comparable eQTL genes in the population, and several of the genes have established relationships with complex disease, indicating a potential for rare variants to further influence genetic risk.
In addition, as genome-interpretation approaches are becoming increasingly informed by diverse noncoding genome annotation,24–26 genome and transcriptome analysis within a single large family provides unique insight into the predictive power of diverse noncoding annotation for rare variants. In our study, we demonstrated that the combination of variant location, epigenomic information, and evolutionary constraint is considerably more informative for predicting the impact of rare noncoding variants than for predicting the impact of common variants. Likewise, we observed equivalent increases in predictive strength for rare splicing variants. This suggests that many rare noncoding variants are likely to be interpretable via existing noncoding annotation and supports their more routine integration in rare-variant association studies.