Secondly, As the interaction predictions based on ortholog are more reliable, among the LR generated from 4 different ortholog interactomes, Caenorhabditis elegance has the lowest LR with the score 92, we used 92 as the cutoff for the medium confident interactions in our network, the 16429 interactive pairs with 5363 proteins.
Confident interactions revealed low-throughput biochemical are shown in Network S2.
As an example of biological hypothesis generation using the DEVEL networks, we investigated the most confident interactions predicted for a specific protein, AtPPT2 (AT3G01550) within two development stages.
Since ChEMBL provides the predicted interactions (not approved yet), we only selected the most confident interactions with the score of 1 under the cut-off of 1 μM [ 5].
Although the connected approach recovers 87.6% of the true positives, the TPR/FPR ratio of the MST approach cannot be reached even if only the most confident interactions in the connected network are considered (see Supplementary Fig. 1).
The network G Kim was compared to G KroganHigh, the set of high-confidence interactions reported by Krogan et al., and to G K r o g a n H i g h T o p, a subset of G KroganHigh consisting of the 164 (to compare against G Kim) most confident interactions they reported.
An initial template functional network was generated from the confident interactions obtained by merging S. meliloti functional genomics data hosted by the PROLINKS [ 13] and STRING [ 14] databases (see methods) (see Fig. 1 for a schematic and full description of the approach).
The cardiac-specific PPI network was built by retrieving relatively high confident protein interactions (confidence score: 0.7) from STRING and gene expression data for MI Wistar rat hearts from GEO (accession number: GDS808).
We found that integrating two data (co-expression and predicted regulatory interactions) enhanced the number of highly confident regulatory interactions by over 10% compared with using single data.
In this way, we obtain a total of 262 883 confident negative interactions for all 70 SH3 domains (the full list of positive and negative interaction data along with the class balance is given in Supplementary Table S2).
