Mapping resolution is determined by the number of meioses (number of recombinants) available. Rather than producing a large number of offspring, which is laborious and expensive, it is much more efficient to make use of historical recombinations. Utilising such linkage disequilibrium (LD) information has been successfully applied to the fine-scale mapping of human diseases. The idea behind LD mapping is that if a marker is sufficiently closely linked to the disease gene, there have been no recombinations between the original marker allele and the mutation that introduced the disease. Hence, an association between a marker allele and the disease, i.e. LD, indicates that the disease is close to this marker. Thus, LD mapping uses all historical recombination events that have occurred since the original mutation, while linkage analysis (such as interval mapping) only utilises recombinations within the data set under study. LD has also been suggested for high resolution mapping of QTL found in a genome wide scan. Recently, the combination of LD and linkage analysis information, has been shown to be even more powerful for fine scale mapping of QTL, narrowing down gene positions to approximately 1 cM. The current project has built on this expertise, extended and tested the statistical method for practical use.
In order for the method to be used efficiently on real data in the current project, several functionalities have been improved. Methods were developed to calculate IBD probabilities taking into account the family structure and the mixture of breeds included in the study. Functionality was added to the IBD-DMU package, allowing for haplotype reconstruction, missing markers and null alleles. When possible a null-allele is considered a normal allele and given an identifier, so that even in the presence of null-alleles some more information with regard to haplotype IBD probabilities can be gleaned.
Further the method used for the construction of the IBD matrix needed verification. The finemapping methods used are based on a matrix of IBD probabilities conditional on the haplotype similarities. These are based on pairwise comparisons of all founder haplotypes and the resulting matrix is therefore most often non-positive definite and cannot be inverted as required. Therefore the matrix must be manipulated somehow, for which several strategies can be used. Another challenge is the large number of haplotypes to be considered. The size of the matrix changes quadratically with the number of haplotypes per QTL locus. Especially in large experiments or in a commercial setting LDLA analysis will demand more computer resources then are readily available. A number of ways to cluster haplotypes and bend the resulting LD matrix or ignore off diagonals between clusters were programmed. An average-linkage hierarchical cluster analysis (Anderberg, 1973) was used to cluster haplotypes according to IBD probabilities in a similarity clustering.