Table 1.

Cancer protein data, comparison of methods

Target	Akt	p70S6K	S6	p53
G.K. Kinetics	4	3	1	8
Lin. Bayes	10	9	15	32
Lin. Bayes Adj.	14	8	8	14
Lin. Lasso	NA	8	NA	NA
Lin. Lasso Adj.	NA	12	NA	NA
Total No of candidates	36	37	36	37

The proposed method was compared with linear approaches using reverse-phase protein array data for nodes whose regulation has been extensively studied in the literature. Each method ranked potential regulators among candidate proteins; here, we display the rank assigned to the known kinase, using each of the five methods. For example, Figure 3b shows such an analysis for the target node (i.e. network child) S6, where G.K. Kinetics ranked the known kinase p70S6K 1st out of a total of 36 candidates. High rank indicates that the known kinase is correctly highlighted in the analysis; the highest-ranked result is highlighted in bold for each target node. Here, we show the rank assigned to the known kinase for each of the target nodes Akt, p70S6k, S6 and p53. (‘NA’ indicates that the known kinase received zero weight. Alternative phospho-forms of the target were excluded as candidates for Akt and S6, so that there were 36 candidates rather than 37. Here, we present results obtained using data from cell lines of basal subtype; luminal results are shown in Supplementary Information.)