Fig. 2

Flowchart of GraphDTI. The input to GraphDTI comprises four feature vectors, A a local network environment for the target protein encoded with Graph2vec, B a drug chemical structure encoded with Mol2vec, C a target protein sequence encoded by ProtVec, and D the structural and physicochemical properties of a binding site encoded with Bionoi-AE. E A feature selection is employed prior to the input layer in order to reduce the dimensionality of the feature vector. F) An input layer concatenating network environment (blue), drug (yellow), protein (red), and binding site (green) feature vectors. G Two hidden layers with selected connections for three neurons colored in dark gray. H An output layer consisting of two neurons to estimate the probability of the drug-target interaction (P—positive, N—negative)