Network-based prediction approach for cancer-specific driver missense mutations using a graph neural network

BMC Bioinformatics

Table 1 Details of three knowledge graphs

Graph	Node	Edge	Edge type	Data source
Cancer pathway	200	870	23	PathwayMapper [8]
Molecular interaction	30,899	3,672,040	13	Pathway commons [10]
Cancer pathway + molecular interaction	30,923	3,672,910	36	PathwayMapper [8], Pathway commons [10]

Number of nodes, number of edges, and types of edges in the three graphs. In the “Cancer pathway” graph, each node represents a protein, complex, and process, such as “apoptosis”, and the edge type represents a type of relationship between them, such as “Activates” and “Inhibits.” In the “Molecular interaction” graph, each node represents a protein and a small chemical compound, and the edge type represents a type of relationship between them defined by Pathway Commons, such as “controls-phosphorylation-of” and “controls-state-change-of”

ISSN: 1471-2105