Table 1.
Composition of the CORUM database.
Figure 1.
A network representation of the CORUM database.
Nodes represent complexes and are ordered by number of unique components (shown as number next to groups). Edges denote shared components between complexes. The number of shared components is reflected in the colour (from yellow (few) to red (many) shared components) as well as in the line width. The large, highly overlapping complexes in the first row are mainly modules of the ribosome (6 out of 12) and spliceosome (3 out of 12). Other large complexes include RNA polymerase, respiratory chain complex and the proteasome. The group of complexes with only 1 member are homo-dimers.
Figure 2.
Coefficients of gene expression variation (CV), defined as standard deviation normalised to expression mean, vary for CORUM and non-CORUM genes.
A) Effects of resolution and dynamic range of expression arrays on CVs. The measurable variation in gene expression is limited by the sensitivity of the employed array technology. Genes which are expressed at extremely low levels, or not expressed at all, cluster in the low expression/low CV region. Shown in grey are genes which were excluded from further calculations (standard deviation ). B) CORUM genes have significantly smaller CVs than non-CORUM genes. Outliers beyond
are not shown. C) Large CORUM complexes exhibit lower average CVs of their members.
Figure 3.
Difference between deletion (white) and duplication (black) variations in HapMap individuals.
The histograms show the ratio of average expression levels between individuals with and without the CNV for all genes inside a CNV region. The shift between the two distributions is significantly larger than would be expected by chance (MWU: ).
Figure 4.
Ratio of WGTP array hybridisation intensity over relative expression level for four example genes.
A) L1CAM and B) UTX. The increase in expression as a result of the copy-number increase in females is clearly visible for UTX which is known to escape X-inactivation. C) and D) Examples of autosomal genes with common CNV polymorphisms. Red crosses denote individuals in which a deletion phenotype has been called by Redon et al., red triangles denote duplications. The plot highlights several potential false negatives with similar expression and hybridisation strength as the called deletions/duplications. Non-CNV related expression variation is substantial.
Figure 5.
Distribution of average Pearson correlation coefficients between all members of known proteins complexes as defined in CORUM (black), and randomly sampled proteins (white, N = 10).
Expression data was taken from the Human Gene Expression Atlas (see Methods).