Figure 1.
Schematic of the structure of long range linkage disequilibria.
Right: Sites A and B are the target of selection or other force that generates disequilibria between them (red arrow). Site A is in disequilibrium with site A′ nearby as the result of shared ancestry of that small segment of chromosome; likewise B and B′ also show short range disequilibrium (blue arrows). Right: A triangle plot of LD. Disequilibria between nearby sites appear as the band along the diagonal. The dash line encloses a “patch” consisting of pairs of widely-separated SNPs that are in LD.
Figure 2.
Schematic of the randomization method used to construct null distributions for patterns of LRLD.
For each randomization, the identity numbers of the haplotypes are permuted randomly (left hand column). The disequilibrium between sites A and B is calculated using the allele at site A from one chromosome and the allele at site B from the following chromosome.
Table 1.
Results for the test statistics pDmax and nP.
Figure 3.
Patches of LRLD on Chromosome 1.
Left: The circle plot represents patches by lines that connect the chromosome blocks involved. The most extreme value of pD in each patch is represented by the color of the segment. The line segments on the outside of the circle show regions identified in Akey's [36] catalogue of genomic targets of positive selection. Right: The triangle plot shows the patches as circles whose size is scaled to the value of the most extreme value of pD in that patch.
Figure 4.
Patches of LRLD on Chromosome 12.
Figure 5.
Patches of LRLD on autosomes 2 to 11.
Figure 6.
Patches of LRLD on Chromosome 12 to 22.
Table 2.
The number of patches that appear in the Akey's [36] catalogue of putative targets of selection.
Table 3.
The effects of the values of the nuisance parameters rmin and n1 on the number of chromosomes that show significant values for the test statistics pDmax and nP at the nominal p<0.05 level (without a Bonferroni correction).
Figure 7.
Scatter plot of the maximum value of −ln(pD) in every patch across all chromosomes vs the distance r between the patches.
Recombination is expected to make this correlation negative.
Table 4.
Spearman rank correlations between −ln(pD) and distance, and between patch density and distance.
Figure 8.
Sved's [17] measure for the correlation of heterozygosity between blocks of 50 SNPs as a function of the recombination rate between the blocks.
The squares show the genome-wide average for the Yoruba population (calculated by Sved). The circles pertain to the 10 cM region of Chromosome 8 that has the highest density of LRLD patches identified by our method. The difference is highly significant (p<10−3).