Fig 1.
The process of constructing a GO-PPI network.
Table 1.
The detailed information of four PPI datasets.
Table 2.
The number of interactions and the portions of essential proteins under the original PPI network and GO-PPI network for the YDIP dataset.
Table 3.
The portions of interactions under the original PPI network and GO-PPI network for the YDIP dataset (BP).
Table 4.
The numbers of essential proteins detected by the six centrality methods under different strategies for the YDIP dataset (top 600).
Table 5.
The number of essential proteins detected by the six centrality methods under different strategies for the new DIP dataset (top 600).
Table 6.
The number of essential proteins detected by the six centrality methods under different strategies for the Krogan dataset (top 600).
Table 7.
The number of essential proteins detected by the six centrality methods under different strategies for the Krogan Extended dataset (top 600).
Fig 2.
The numbers of key proteins predicted correctly under the original PPI network and reconstructed GO-PPI network for the YDIP dataset.
Fig 3.
Prediction precision of six centrality methods under the original PPI network and reconstructed GO-PPI network for the YDIP dataset.
Fig 4.
The comparison of ROC curves for original new DIP PPI network and Wang-BP (the Wang method under the BP subontology) PPI network under (left) EC method, (middle) SC method and (right) aveNC method.
Fig 5.
The comparison of ROC curves for original YDIP PPI network and Resink-BP (the Resink method under the BP subontology) PPI network under (left) BC method, (middle) DC method and (right) aveNC method.
Fig 6.
The numbers of correctly identified key proteins among the top 100, 200, 300, 400, 500, and 600 candidates for different method vs the threshold value.