Rare variant association studies: considerations, challenges and opportunities

Abstract

Genome-wide association studies (GWASs) have successfully uncovered thousands of robust associations between common variants and complex traits and diseases. Despite these successes, much of the heritability of these traits remains unexplained. Because low-frequency and rare variants are not tagged by conventional genome-wide genotyping arrays, they may represent an important and understudied component of complex trait genetics. In contrast to common variant GWASs, there are many different types of study designs, assays and analytic techniques that can be utilized for rare variant association studies (RVASs). In this review, we briefly present the different technologies available to identify rare genetic variants, including novel exome arrays. We also compare the different study designs for RVASs and argue that the best design will likely be phenotype-dependent. We discuss the main analytical issues relevant to RVASs, including the different statistical methods that can be used to test genetic associations with rare variants and the various bioinformatic approaches to predicting in silico biological functions for variants. Finally, we describe recent rare variant association findings, highlighting the unexpected conclusion that most rare variants have modest-to-small effect sizes on phenotypic variation. This observation has major implications for our understanding of the genetic architecture of complex traits in the context of the unexplained heritability challenge.

Figure 1

Comparison of power for trios and case–control designs. Power to detect associations for 10,000 cases and 10,000 controls (blue) and 10,000 trios (red) across a range of minor allele frequencies (MAFs). Power was calculated with a significance threshold of P < 0.05, a prevalence of 0.1 and a relative risk of 1.1, using the Genetic Power Calculator tool [112].

Figure 2

Functional annotation of regulatory sequences in the human genome. Genome tracks from the UCSC Genome Browser. CXCL2 (blue) encodes a chemokine produced by activated monocytes and neutrophils at sites of inflammation. Single nucleotide polymorphisms (SNPs; rs546829 and rs1371799, green) are associated with monocyte count by a genome-wide association study. The red box upstream of CXCL2 includes a predicted enhancer identified in monocytes by FANTOM5 (black rectangles). FANTOM5 did not annotate an enhancer in hepatocytes, a less relevant cell type for CXCL2. Using histone tail modification information, ENCODE predicted strong enhancers (orange) at the same position in erythroleukemic (K562) and endothelial (HUVEC) cells. Chr, chromosome; hESC, human embryonic stem cell; HMM, hidden Markov model; kb, kilobases.