Learning from biomedical linked data to suggest valid pharmacogenes

Table 1 Example of a feature matrix generated from linked data

ID	Gene attribute	Phenotype	Drug attribute	G-D link	G-P link	D-P link	Class
PA7360-PA131301952	Signal transduction	C0007131	L01XE2	Antagonist	clinvar: so_0001575	sider: indication	1
PA7360-PA131301952	Immune system	C0007131	L01XE2	Antagonist	clinvar: so_0001575	sider: indication	1
PA7360-PA131301952	Signal transduction	C0007131	L01XE2	Antagonist	clinvar: so_0001619	sider: indication	1
PA7360-PA131301952	Immune system	C0007131	L01XE2	Antagonist	clinvar: so_0001519	sider: indication	1

All the instances (e.g., lines) describe the same gene–drug relationships (EGFR–Gefitinib), which is associated in PharmGKB with a high level of evidence (Class=1). Figure 3 shows some of the data associated with this relationships in linked data. Values are extracted from the graph and are encoded in various manner. For example, values of phenotypes are UMLS CUI, values of drug attributes are ATC codes

ISSN: 2041-1480