Complex trait rare variant association studies using sequence data are being widely performed. Analyzing rare variants individually is extremely underpowered; therefore many powerful rare variant association methods have been developed, which are based upon jointly analyzing multiple variants within a region which is usually a gene. After an association is identified, it is also important to estimate genetic parameters of interest and quantify the proportion of heritability explained by the gene. A drawback of rare variant association methods is that it is not possible to tease apart causal from non-causal variants. Consequently, the causative-variant-effect is not estimable. We describe how to efficiently estimate the genetic-average-effect. Due to the presence of non-causal variants, genetic variance explained by the genetic-average-effect will be underestimated but provides a lower bound for the true underlying locus genetic variance. An additional problem is due to the winner’s cure the naïve estimator can be seriously inflated. The bias is quantified and it is shown that even for poorly powered studies a boot-sample-split procedure can be used to greatly reduce the bias of the genetic estimates. Not only are these methods vital for estimating the amount of missing heritability due to rare variants, but they are also important for designing replication studies and risk prediction.
Back to Workshop IV: Coancestry, Association, and Population Genomics