Network Edge Reconstruction Strategy
We performed a random walk with resistance (RWS) on the biology network, and derived a modified network by adding and removing some edges. The results shows that the adding edges in the reconstructed network have higher biological relevance than in the original network.
However, we just used the simplest strategy to keep the same amount of connection as the original network, which is by choosing the potential connections with the highest similarities.
Here, we will discuss some advanced strategy to
reconstruct the network, and hopefully the new network will show a much
better biological performance than the original bio-network.
1, Original Paper and the Supplementary data can be found here and here.
2, We use Yeast and Human PPI network to test.
-
Yeast
-
Krogan's data download (Local mirror)
-
local download for the Original network, names, full correlation matrix
-
-
-
BioGrid for verification download
-
Choose "Saccharomyces_cerevisiae" for (human language) yeast data
-
Column A and B will be two proteins that have a connection
-
Column G for different bio-experimental technology
-
-
Human
3, Set up experimental environment for Krogan's data
-
convert all the local data of Krogan dataset to txt file, and write python code to read all the files appropriately
-
Process the BioGrid data, separate the PPI data by different bio-experimental technology.
-
Try to plot the performance curve for different correlation cutoff.