Figure 1.
Spc and NetView analysis of the simulated data set.
(A) Population structure used for creating the simulated data (adapted from Lawson et al. [19]). (B) Spc tree of clusters representing the grouping of individuals with k-NN = 10. The individuals have been separated into 5 clusters, representing the three main populations and the additional existence of two sub-populations (PopA1 and PopA2, PopB1 and PopB2). Each cluster is represented by a box; with Y axis positions indicating the stability of each cluster, whilst the X-axis positions are indicating the proximity between clusters. (C) High-definition network visualization (NetView) of the simulated population structure. Each individual is represented by a node; with the different shades denote the sample origin. The thickness of edges varies in proportion to the genetic distance and has been used to visualize individual relationships within and between populations. The node size varies in proportion to the numbers of edges per node, and illustrates how well each individual is connected within the population.
Figure 2.
PCA scatter plots of the simulated data set.
Projection of individuals from 5 populations onto a two dimensional (X,Y) subspace of four PCs. The panels A to D show pair wise comparison of PC combinations. Each individual is represented by a datum point. Each sub-population is denoted by a separate colour. The variation captured by each PC is indicated in parenthesis next to the axis label.
Figure 3.
Cluster assignment of the simulated population data following analysis by Admixture using 2–5 clusters (K).
Individuals are presented by a single vertical column divided into K colours. Each colour represents one cluster, and the length of the coloured segment corresponds to the individuals estimated proportion of membership in that cluster. For each K, 10 iterations were performed. The panels A to D represent the cluster patterns at K = 2 to 5.
Figure 4.
Spc and NetView analysis of human HapMap reference population after removal of closely related individuals.
(A) Spc tree of clusters representing the grouping of 1,159 unrelated individuals. All individuals have been separated into 11 clusters, representing 9 distinct populations and the existence of sub-structures within GIH and MKK samples. (B) NetView of the 1,159 assumed unrelated individuals. The topology of the network highlights the sub-structures within GIH, MXL and MKK and reveals a close relationship between CEU and TSI as well as between ASW and MKK. The identified outliers and key individuals of the population are indicated by their HapMap ID.
Figure 5.
Alternative (organic) NetView of populations with evidence of internal sub-structures.
Organic visualization style of (A) ASW/MKK and (B) TSI/CEU as implemented in software Cytoscape [59]. The network structure of this visualization highlights the existence of sub-structures and clearly identifies cross-linking individuals.
Figure 6.
Spc and NetView analysis of the Bovine HapMap data.
(A) Spc tree of clusters representing the grouping of 477 animals represented in the Bovine HapMap data set [31]. The animals have been allocated into 19 clusters, representing 18 out of 19 breeds and the existence of sub-structures within JER (JER_1 and JER_2), and a merged Angus cluster (ANG and RGU). (B) NetView of 477 bovine HapMap samples from Bos taurus, Bos indicus and admixed origins. The topology of the network reflects the genetic relatedness between cattle breeds and reveals sub-structures within LMS, SHK and ANG cluster.
Table 1.
Overall comparison of the three different approaches currently applied to study genome-wide population structures.