Hierarchical clustering analysis of blood plasma lipidomics profiles from mono- and dizygotic twin families

Abstract

Twin and family studies are typically used to elucidate the relative contribution of genetic and environmental variation to phenotypic variation. Here, we apply a quantitative genetic method based on hierarchical clustering, to blood plasma lipidomics data obtained in a healthy cohort consisting of 37 monozygotic and 28 dizygotic twin pairs, and 52 of their biological nontwin siblings. Such data are informative of the concentrations of a wide range of lipids in the studied blood samples. An important advantage of hierarchical clustering is that it can be applied to a high-dimensional 'omics' type data, whereas the use of many other quantitative genetic methods for analysis of such data is hampered by the large number of correlated variables. For this study we combined two lipidomics data sets, originating from two different measurement blocks, which we corrected for block effects by 'quantile equating'. In the analysis of the combined data, average similarities of lipidomics profiles were highest between monozygotic (MZ) cotwins, and became progressively lower between dizygotic (DZ) cotwins, among sex-matched nontwin siblings and among sex-matched unrelated participants, respectively. Our results suggest that (1) shared genetic background, shared environment, and similar age contribute to similarities in blood plasma lipidomics profiles among individuals; and (2) that the power of quantitative genetic analyses is enhanced by quantile equating and combination of data sets obtained in different measurement blocks.

Figure 1

Heatmaps of Euclidean distances between objects, and associated hierarchical clustering dendrograms for combined B1–B2 data set before and after quantile equating. B1 and B2 data sets were combined by concatenation with the variables (lipids) as the shared mode. (a) Before quantile equating; (b) After equating. In this figure, individual objects are labeled by two color codes: the first color encodes the gender of the participant of whom the sample was obtained (red for females and blue for males). DZ female and DZ male twins are indicated with pink and light blue, respectively. The second color encodes the block in which the sample of this participant was measured (white for B1 and black for B2).

Figure 2

Box-whisker plots showing distributions of Euclidean distances between MZ cotwins (N=37 distances), between DZ cotwins (N=28 distances), among sex-matched nontwin siblings (N=66 distances), and among sex-matched unrelated participants (N=8203 distances) in the combined equated B1–B2 data set. The observations indicated with a plus sign in case of the unrelated participants illustrate the slight skewness of the distribution of the Euclidean distances among all participants.

Figure 3

Results of node analyses with respect to permutation-based chance distributions. (a) MZ cotwins; (b) DZ cotwins; (c) Sex-matched nontwin siblings. Numbers of nodes separating cotwins or nontwin siblings increase from left to right in each panel. For each number of branching points, from bottom to top the number of twin or nontwin sibling pairs separated by that particular number of branching points in the permutation tests is displayed by gray bars. Black dots indicate the number of observations given the original ordering of labels along the leaves of the dendrogram as in Figure 1b and Supplementary Figure S2. The depicted chance distributions were created by combination of the results from all (ie, 100) sets of 10 000 permutations. Asterisks indicate average P-values <0.05 (see Supplementary Table S2).