Table 1 Pairwise comparison of the methods

From: Predicting drug combination response surfaces

View full size image

  1. The table reports average percentages over all cell lines. How often did the method on row give a better prediction for a surface, w.r.t. mean squared error, than the method on column; rows with high values (darker colouring; similarly columns with low values with lighter colouring) indicate better performance for the method. The two comboKR versions with original and normalised surface kernels use the same candidate sets, and so their predictions might be identical, resulting in ties in the rankings. Tie counts are not included in the tables, so cells marked with asterisk do not add to 100%.
Back to article page