Figure 2
From: Query-based biclustering of gene expression data using Probabilistic Relational Models
Effect of the seed set quality on the quality of the retrieved biclusters. (x-axis) STD-across/STD-within conditions of the seeds, (y-axis) STD-within/STD-across (log2 transformed) of the retrieved biclusters. Bicluster results of respectively Pro Bic (blue) – QDB (green) – ISA (red) are subdivided in the following groups (indicated by the respective markers): ‘x’ seed sets that give rise to biclusters that contain the full seed set or part of the seed set together with additional genes, ‘+’ seeds that give rise to biclusters that do not longer contain the seed sets (drift away biclusters), ‘o’ seed sets that give rise to ‘empty’ biclusters or biclusters that contain seed genes only. Only the results of 127 seed sets of which the seed quality could be calculated (seed set should contain more than one seed gene), are displayed. Full lines display the overall trend.