Linkage disequilibrium is the non-random allelic association between different loci. Recently, it has been advocated as a powerful tool to refine gene location estimates, thanks to its ability to incorporate the effects of a large number of past generations of recombination. In localizing mendelian disease genes in human populatiions, this approach has already been succesfull. But more research work is still needed to determine its utility for fine mapping Quantitative Trait Loci (QTL) in animal populations. My talk gives some new insights on this question.

I first present an interval mapping method for QTL. Following Xiong and Guo (1997), we use a maximum likehood approach and model the joint history of the quantitative trait locus and two flanking markers as a stochastic process. Since we deal with QTL, we can't assume as they do that one of the trait alleles is rare, and this brings us to derive the expected haplotype frequencies in a slightly different way. We compare our method with other existing ones under several simulation scenarios. Mean square errors of the QTL location estimates are significantly smaller than those obatined with Abdallah et al (2004)'s method, and similar to those obtained by Meuwissen and Goddart (2000). And our method is much faster than this last one.

Finally I propose a different way of evaluating the likehood above, based on a diffusion approximation of the Wright Fisher model. It gives a more precise estimate of the likehood, taking account of the stochasticity of haplotype frequency evolution over generations. It applies to many kinds of populations.