Figure 2.

The iDSN and contributions of various data types

The 593 drugs in the iDSN are grouped into 32 clusters. Each node represents a drug, and node color signifies cluster assignment of the drug. Edges are weighted according to similarity score. Edges are color coded according to the major contributing data types, as defined in the pie chart on the right. Three types of data, including chemical structure, drug targets, and side effects, are used alone or in various combinations to evaluate their respective contributions. Clusters 3 and 28 are zoomed in as examples for better illustration.