Relationships decomposed into components. “Eye diagram” showing the top 10 significant CCA components ordered by correlation from top to bottom (middle), VolSurf descriptors (left), and top gene sets (right). The CCA components are shown as circles, with numbers indicating the decreasing order of canonical correlation and letters A and B indicating subcomponents (A: positive canonical score, B: negative canonical score). The widths of the curves from the components to VolSurf descriptors and gene sets indicate the strength of the corresponding associations. For VolSurf descriptors the subcomponent-specific activity is shown, whereas for the gene sets we show the overall activity in the component. For an example compound, VolSurf fields are illustrated in the top-left corner while three gene sets are listed along with their five most significant genes in the top right corner.