Figure 2

Study 2 undirected edge posterior probabilities. For the second study, (a), (b), (c) and (d) are representations of the undirected edge posterior probabilities. These are computed for the composite of all the replicates, C⁽²⁾, and for the individual replicates, $R_1^{\left(2\right)}$, $R_2^{\left(2\right)}$ and $R_3^{\left(2\right)}$. Blocks of vertices $\left\{p_1^{\left(2\right)},p_2^{\left(2\right)},p_3^{\left(2\right)}\right\}$, $\left\{p_4^{\left(2\right)},p_5^{\left(2\right)}\right\}$, $\left\{p_6^{\left(2\right)}\right\}$, and $\left\{p_7^{\left(2\right)},p_8^{\left(2\right)},p_9^{\left(2\right)}\right\}$ are highly correlated, ρ = 0.9, by the generating signal (3). In the diagrams proteins $p_1^{\left(2\right)},\dots ,p_9^{\left(2\right)}$ are represented by A, . . . , I for ease of reading. As in the first study, the composite has high posterior probabilities for all edges with at least moderate network partial correlations. Overall, the smaller blocks were not identified as well as the triple blocks. The signal was not identified as well for $R_1^{\left(2\right)}$.