Constructing a drug–drug functional similarity network based on the D/G bipartite network

Furthermore, we constructed a drug–drug functional similarity network, in which the edge weights represent the similarity scores between drugs. To represent the drug–drug network, we defined an adjacency matrix A:

where W is the adjacency matrix of the D/G bipartite network. Thus, the elements of A correspond to the functional similarity scores (see equation (3)) between drugs in the drug–drug similarity network. We assign the value of 0 to the diagonal element of matrix A, indicating self-links were removed. After these steps, a drug functional similarity network under a specific disease context is then constructed, which comprises a total of 1,289 drug nodes and 322,250 edges. This process was previously used to construct a cell–cell crosstalk network [27].