e -CCC-Biclusters extended to "jump over" missing values and allow anticorrelation. This figure shows: (Top) Generalized suffix tree used by e-CCC-Biclustering extended to "jump over" missing values and extract e-CCC-Biclusters with sign-changes when applied to the transformed matrix in Figure 9. The circles labeled with B1, B2 and B3 identify the node-occurrences of the three maximal 1-CCC-Biclusters discovered when e = 1, q
= 3 and q
= 2; (Bottom) Maximal 1-CCC-Biclusters corresponding, respectively, to the valid models m = [D3 D4] (B1), m = [D3 U4 D5] and m-1 = [U3 D4 U5] (B2), and m = [U4 U5] and m-1 = [D4 D5] (B3).