(All) Statistics of the sign-inference process on the regulatory network of E. coli from partial expression profiles. The setting is similar to the one used in Fig. 6, except for the cardinal of the expression profiles (N is fixed), and for the variable on X-axis which represents the percentage of missing values in the expression profiles. The continuous line corresponds to the theoretical prediction Mi = - d * f * Mtotal; where is the number of inferred interactions from complete expression profiles, d is the number of interaction signs no longer inferred when a node is not observed, f is the fraction of unobserved nodes, and Mtotal is the total number of nodes. (Left) Statistics for the whole network; we used 30 sets of artificial expression profiles (N = 30). We estimated d = 0.35, meaning that on average we lose one interaction sign for about 2.9 missing values in the profiles. (Middle) Statistics for the core network (N = 30). We estimated d = 0.43; the core of the network, however, is more sensitive to missing data. (Right) Statistics for the core network (N = 200). We estimated d = 0.74; hence, increasing the number of expression profiles increases the sensitivity to missing data.
Veber et al. BMC Bioinformatics 2008 9:228 doi:10.1186/1471-2105-9-228