Deciphering the genetic architecture of a multivariate phenotype.

A heritable multivariate quantitative phenotype comprises several correlated component phenotypes that are usually pleiotropically controlled by a set of major loci and environmental factors. One approach to decipher the genetic architecture of a multivariate phenotype, in particular to map the underlying loci, is to reduce the dimensionality of the data by means of a data reduction technique, such as principal component analysis. The extracted principal components are then analyzed in conjunction with marker data to map the underlying loci. We have examined the efficiency of this approach with and without taking into account the correlation structure of the multivariate phenotype when extracting principal components. We have assumed that genome-wide scan data on sibpairs are available for low-density (widely spaced) and high-density markers. Using extensive simulations, based on three models of the multivariate phenotype, we have shown that although ignoring the correlation structure of the multivariate phenotype does not have any serious impact on the efficiency of mapping the underlying trait loci in wide marker intervals, there is a significant adverse effect of this practice for fine-mapping. We, therefore, recommend that the correlation structure of the multivariate phenotype be carefully examined to decide on the strategy of extracting principal components for deciphering the genetic architecture of the multivariate phenotype.