Genetic Analysis of Transformed Phenotypes

Nicolo Fusi, Christoph Lippert, Neil D. Lawrence, Oliver Stegle
(Submitted on 21 Feb 2014)

Linear mixed models (LMMs) are a powerful and established tool for studying the genetics of phenotypic variation. A limiting assumption of LMMs is that the phenotype is Gaussian distributed under the model, a requirement that rarely holds in practice. Since violations of this assumption can lead to false conclusions and losses in power, it’s common practice to pre-process the phenotypic values, for instance by applying logarithmic transformations. Unfortunately, these are not appropriate in every situation, and choosing a “good” transformation is in general challenging and subjective. Here, we present an extension of the LMM that estimates an optimal transformation from the data. We show in extensive simulations and real data from human, mouse and yeast that application of these optimal transformations leads to increased power in genome-wide association studies and higher accuracy in heritability estimates and phenotype predictions.