A heat map representation of the spiked-in correlation network in the simulated data. The heatmap is a two dimensional projection of a four dimensional matrix, 15 × 15 genes × 3 × 3 genotypes. Here the 3 × 3 cross-genotype blocks are nested within each gene block. As a self-correlation matrix, the column IDs are identical to the row IDs. The left panel shows the two sub-networks that were used to drive the simulation, one involving CDH1 and CDH10, the second involving CDH19, PCDH1, PCDH10, and PCDH17. PCHD19 interacted with several genes, but only under certain genotype configurations. This matrix also implies other high order dependencies that are not well shown in this form, but can be observed by tracing from a significant value in a cell, to any other significant value for another gene that occurs in either the same row or column. The number of steps along which such a chain may be followed, defines the number of interacting factors. The correlation matrix re-derived from the output of the simulation (right panel) includes both the spiked-in network and stochastic variation from the simulation, as well as the real biological correlations across genes.
Bartlett et al. BMC Bioinformatics 2012 13(Suppl 8):S8 doi:10.1186/1471-2105-13-S8-S8