Reconstructing cancer drug response networks using multitask learning

BMC Systems Biology

Table 1 Comparison of different gene and network analysis methods for the reconstruction of drug response networks

Control	MT	Cell 1	Cell 2	MT-Diff	Diff-Cell1	Diff-Cell2
CGC	13.33	8.5	8.66	6.33	4.66	6.5
GO	72.33	33.66	41.5	47.66	43.16	29.5
Oncogenic	7.66	3	3.33	10.33	10.33	4.83
Breast & Prostate
CGC	14.66	8.33	10	2.66	3.33	1.83
GO	77.5	70.66	64.83	18.16	25.66	18.33
Oncogenic	8.66	4.33	5.16	2.66	2.5	2.5
Prostate
CGC	15	10.16	10.33	3.33	2.33	3.83
GO	82.33	85.83	88.66	23	26.83	18.5
Oncogenic	11	8.33	7.66	3	4.5	3.16

Values for each gene set and learning method denote the average number of genes (across six drugs) selected by each method which are also contained in the corresponding validation set. MT: multi-task, Cell 1, Cell 2: single task analysis (cell type based) for the two cells. “Diff” columns show genes selected only by differential expression (DE); MT-Diff: DE set for the two cell types, selecting genes that are differentially expressed in both cells, Diff-Cell1/2: cell type specific DE set

ISSN: 1752-0509