Eric Lander and colleagues report an analysis offering considerations for the design of rare variant association studies (Proc. Natl. Acad. Sci. USA 111, E455–E464, 2014). They consider a binary trait, exemplifying such a disease phenotype, analyzed with burden tests and use a two-class model in which they assume that alleles are either null, abolishing gene function, or neutral, not accounting for intermediate effects. Their simulations estimate the sample sizes needed to detect rare variant associations across a range of study designs, demonstrating a strong dependence on the mutation rate, selection coefficient and effect sizes for null alleles in a gene. For rare variants in coding regions, their first strategy is to consider only disruptive variants in each gene. They suggest that missense alleles may also be included to increase power, but care should be taken to filter using optimally selected frequency thresholds and experimental or computational predictions. They also consider the benefits of studying isolated populations with recent bottlenecks. For noncoding regions, the authors note that a detailed understanding of functionally relevant regulatory sequences for each gene is needed. For current rare variant association studies, they recommend the use of exome, rather than whole-genome, sequencing to maximize power to detect association.