We consider PPIs and L2-PPIs as two separate data sources.
We consider PPIs with a BP score over 0.70 to be supported by evidence of shared biological process.
We collapsed the PSI-MI descriptions of method to the high-level categories presented earlier, and consider PPIs annotated as identified by more than one high-level method to be supported by experimental detection methods.
Each pair of matched interactions defines a candidate transformation that can superimpose the considered PPI upon the pivot.
The reasons for considering PPI therapy to be optimized were recorded by physicians (multiple response possible); the 'maximum dose' had been reached in 86% of patients, 46% of patients had tried a different brand with no improvement, improvement by changing PPI was considered unlikely due to symptom pattern in 29% and PPI therapy was considered optimized due to other reasons in 9% of patients.
Specifically, we considered PPI network features including degree, clustering coefficient and the length of the shortest path to a cancer gene, sequence features including gene and protein lengths, and conservation features including Ka and Ka/Ks ratio.
We also considered PPIs which were generated at least one time in a small-scale manner.
After considering PPIs, it is natural to move to a higher degree of organization of the living matter, i.e. biological pathways.
PPI latent graph: In our ensemble model, we consider the PPI network as a latent graph, and construct the adjacency matrix E(1) of the PPI latent graph as follows E i j 1 = E i, j where E i, j) = 1 if node v i and node v j are connected in the PPI network, and E i, j) = 0 otherwise.
As for the lack of PPI use before ARS, one practical implication could be that surgeons might consider checking PPI compliance by pill count, enquire about prescription data or apply other measures before performing ARS.
