Figure 1.
Comparison of linkage disequilibrium visualization methods applied to the Daly et al. dataset [12].
a) triangular heat map of D′ = LOD with LD blocks computed using Haploview v4.2, b) recombination rates inferred with PHASE v2.1, c) textile plot and d) forest of hierarchical latent class models displayed using Tulip. For each layer l, the median of r2 values (resp. distances) is computed over all pairs of variables having their lowest common ancestor in layer l. The magnified node in subfigured) shows the multilocus linkage disequilibrium measure relative to the latent variable thus highlighted. Dotted lines highlight common trends between the four methods. Association signal is visualized through the length of an edge linking a SNP to an additional node.
Figure 2.
Information of the level of SNP lowest common ancestors.
Relation between the level of SNP lowest common ancestors and a) the median of r2 values and b) the median of distances computed over the corresponding SNP pairs. For instance, the left boxplot in a) concerns all pairs of SNPs with lowest common ancestors belonging to layer 1. Over these SNP pairs, the median of r2 values is 0∶91.
Figure 3.
Long-range linkage disequilibrium visualization for the region [22 Mb–40 Mb], chromosome 6, surrounding the major histocompatibility complex.
a) chromosome map view built with UCSC Genome Browser, b) triangular heat map of D′ = LOD built by Haploview v4.2, c) textile plot and d) forest of hierarchical latent class models displayed using Tulip. For each layer l, the median of r2 values (resp. distances) is computed over all pairs of variables having their lowest common ancestor in layer l.
Figure 4.
Long-range linkage disequilibrium visualization of 2819 regularly spaced SNPs of chromosome 1, for the African ancestry in Southwest USA (ASW) HapMap population.
A long-range LD region is circled in red.
Figure 5.
Chromosome-wide linkage disequilibrium visualization of chromosome 1.
Navigation through zooming inside the FHLCM graph. Positions of SNPs are displayed inside blue nodes. The multilocus LD strengths relative to the subtrees rooted in the latent variables can be viewed inside red nodes.
Figure 6.
The forest of hierarchical latent class models.
The light shade (blue) indicates the observed variables whereas the dark shade (red) points out the latent variables.
Figure 7.
Illustration of specific key terms: subtree, tree, forest, parent, descendant, ancestor and lowest common ancestor.
See Figure 6 for node nomenclature.