Fine-scale linkage disequilibrium mapping: a comparison of coalescent-based and haplotype-clustering-based methods

Among the various linkage-disequilibrium (LD) fine-mapping methods, two broad classes have received considerable development recently: those based on coalescent theory and those based on haplotype clustering. Using Genetic Analysis Workshop 15 Problem 3 simulated data, the ability of these two classes to localize the causal variation were compared. Our results suggest that a haplotype-clustering-based approach performs favorably, while at the same time requires much less computing than coalescent-based approaches. Further, we observe that 1) when marker density is low, a set of equally spaced single-nucleotide polymorphisms (SNPs) provides better localization than a set of tagging SNPs of equal number; 2) denser sets of SNPs generally lead to better localization, but the benefit diminishes beyond a certain density; 3) larger sample size may do more harm than good when poor selection of markers results in biased LD patterns around the disease locus. These results are explained by how well the set of selected markers jointly approximates the expected LD pattern around a disease locus.