Description
The numerical value of interaction in a protein-compound pair determines the tendency of a
compound (drug candidate) to bind to a target protein, which plays a key role in the initial phase
of the drug discovery process. In the first phase of drug discovery, from a large number of
primary compounds, compounds that have the desired properties against the target protein are
selected and introduced to the next phase. The compounds obtained in this phase are known as
the leading compounds.
The goal in this method is to construct a graph in which each node represents a
protein-compound pair and the edges between the nodes also represent the relationship between
the protein-compound pair. In other words, the goal is to include the information of similar
protein-compound pairs in predicting the interaction between them. In this graph, nodes that are
labeled are the training examples we have, and nodes that are not labeled are our samples for
interaction prediction.
