Skip to main content
Fig. 4 | BMC Bioinformatics

Fig. 4

From: Identification of pan-kinase-family inhibitors using graph convolutional networks to reveal family-sensitive pre-moieties

Fig. 4

Feature encoding of the input compounds. Each compound is encoded by atomic features and a structure graph. Atomic features contain 28 descriptors belonging to six types, including atom types, hybridization, charges, and chemo-properties. For a structure graph, a modified normalized Laplacian matrix (Eq. (4)) was applied as a representation of the compound’s topology information. Padding to 50 atoms with zeros was applied to contain variable numbers of atoms in the input compounds

Back to article page