Many of the proteins identified by our screen with FEZ2 interact with proteins that also interact with each other, increasing therefore the likelihood that these are true interactions of biological relevance.
We find nearly 28% of the true interactions out of 684 can be retrieved at the top 10 rank positions and more than 38% of the interactions can be retrieved at the top 50 rank positions.
Among 7870 predicted interactions by ortholog co-occurrence in operons, 2309 were predicted as true interactions on the basis of phylogenetic profiles [See Additional file 6].
Similarly, estimates of true interaction counts permit tests of hypotheses that classes of proteins are hubs.
Because of the sparsity of the initial data and the presumed low density of the true interaction networks, many pairs of orthology groups contain one or zero interactions between them (they appear white in Fig. 4) and are left-out of the experiment.
In the risk-profiling framework, identification of true interaction effects largely depends on the prediction accuracy of the genetic and environmental factors.
A single positive training set, however, may be biased and not representative of true interaction space.
Even training data that has known or undefined bias would be expected to improve this scoring system by giving representation to another region of true interaction space.
The average value of β was then taken as the value used to calculate the probability of true interaction T vu given the three input features X = (X1, X2, X3), which are 1760.946913, 36.81855608, 6.593051189, and −2014.167331 for β0 to β4, respectively.
According to the logistic distribution, the probability of true interaction T vu given the three input features X = (X1, X2, X3) is calculated based on the formula as follows: (3) Pr T v u ∣ X = 1 1 + exp (− β 0 − ∑ i = 1 3 β i X i ).
