Coherent estimates of genetic effects with missing information

Genetic effect estimates for loci detected in quantitative trait locus (QTL) mapping experiments depend upon two factors. First, they are parameterizations of the genotypic values determined by the model of genetic effects. Second, they are consequently also affected by the regression method used to estimate the genotypic values from the observed marker genotypes and phenotypes. There are two common causes for marker-genotype data to be incomplete in those experiments—missing marker-genotypes and within interval mapping. Different regression methods tend to differ in how this missing information is represented and handled. In this communication we explain why the estimates of genetic effects of QTL obtained using standard regression methods are not coherent with the model of genetic effects and indeed show intrinsic inconsistencies when there is incomplete genotype information. We then describe the interval mapping by imputations (IMI) regression method and prove that it overcomes those problems. A numerical example is used to illustrate the use of IMI and the consequences of using current methods of choice. IMI enables researchers to obtain estimates of genetic effects that are coherent with the model of genetic effects used, despite incomplete genotype information. Furthermore, because IMI allows orthogonal estimation of genetic effects, it shows potential performance advantages for being implemented in QTL mapping tools.