Treatment of the X chromosome in mapping multiple quantitative trait loci

Abstract

Statistical methods to map quantitative trait loci (QTL) often neglect the X chromosome and may focus exclusively on autosomal loci. But the X chromosome often requires special treatment: sex and cross-direction covariates may need to be included to avoid spurious evidence of linkage, and the X chromosome may require a separate significance threshold. In multiple-QTL analyses, including the consideration of epistatic interactions, the X chromosome also requires special care and consideration. We extend a penalized likelihood method for multiple-QTL model selection, to appropriately handle the X chromosome. We examine its performance in simulation and by application to a large eQTL data set. The method has been implemented in the package R/qtl.

Keywords: QTL; X chromosome; model selection.

Figure 1

Simulation results: Type I error and power of these methods across simulated models. Black is the method treating the X chromosome and autosomes the same (XeqA); green is the method treating the X chromosome and autosomes differently (XneA). The estimates and 95% confidence intervals are based on 1,000 simulations. (A) Type I error for main effects. (B) log₂ ratio of the Type I error of main effects for autosomes versus the X chromosome. (C) Type I error for interactions. (D) Power to detect QTL.

Figure 2

LOD profile of output models from XeqA (black) and XneA (green) methods on probe 10003837305. Note that there are two black curves on chromosome 7, because the XeqA method identified two QTL on that chromosome.