In Silico Study of Transcriptome Genetic Variation in Outbred Populations

Abstract

Dissecting the genetic architecture of regulatory elements on a genome-wide basis is now technically feasible. The potential medical and genetical implications of this kind of experiment being very large, it is paramount to assess the reliability and repeatability of the results. This is especially relevant in outbred populations, such as humans, where the genetic architecture is necessarily more complex than in crosses between inbred lines. Here we simulated a chromosome-wide SNP association study using real human microarray data. Our model predicted, as observed, a highly significant clustering of quantitative trait loci (QTL) for gene expression. Importantly, the estimates of QTL positions were often unstable, and a decrease in the number of individuals of 16% resulted in a loss of power of ∼30% and a large shift in the position estimate in ∼30-40% of the remaining significant QTL. We also found that the analysis of two repeated measures of the same mRNA can also result in two QTL that are located far apart. The intrinsic difficulties of analyzing outbred populations should not be underestimated. We anticipate that (many) conflicting results may be collected in the future if whole-genome association studies for mRNA levels are carried out in outbred populations.