Fig. 5
From: Choosing panels of genomics assays using submodular optimization
Performance of panel selection strategies. a Box plots show the distribution of the evaluation metrics over 40 random panels on data from cell type GM12878. The panels of the most frequently performed assays are composed of the top k most frequent assay types available in our data set, where k is the size of the panel. Each evaluation metric is normalized to lie within [0,1] by subtracting the lowest value and dividing by the highest. b Scatter plots between the performance of SSA-past and SSA-future across cell types K562, GM12878, and H1-hESC. Each dot in the plot corresponds to the two variants of SSA for a panel size evaluated using a metric in a cell type. The performance is measured as the fraction of panels that perform worse than the SSA-chosen panel, estimated by comparing to 40 randomly selected panels. SSA submodular selection of assays