A knowledge graph approach to predict and interpret disease-causing gene interactions

BMC Bioinformatics

Table 1 Knowledge graph edge types

Metaedge	Abbreviation	\(\#\) Edges	\(\#\) Sources	\(\#\) Targets
Gene–coexpresses–Gene	GeG	1,338,764	14,940	14,940
Gene–physinteracts–Gene	GpG	329,801	17,062	17,062
Disease–described\(\rightarrow\)Phenotype	DdP	233,175	12,676	10,423
Gene–associated\(\rightarrow\)Phenotype	GaP	209,416	4870	9151
Gene–seqsimilar–Gene	GsG	186,445	12,226	12,226
Gene–associated\(\rightarrow\)BiologicalProcess	GaBP	93,676	16,323	10,570
Gene–associated\(\rightarrow\)CellularComponent	GaCC	58,432	16,978	691
Gene–belongs\(\rightarrow\)ProteinFamily	GbPF	45,454	19,657	11,187
Gene–associated\(\rightarrow\)MolecularFunction	GaMF	43,331	14,540	4042
Gene–hasunit\(\rightarrow\)ProteinDomain	GuPD	41,314	15,828	6636
BiologicalProcess–resembles–BiologicalProcess	BPrBP	33,102	10,811	10,811
Phenotype–resembles–Phenotype	PrP	16,000	7681	7681
Gene–forms\(\rightarrow\)ProteinComplex	GfPC	14,531	4357	3604
MolecularFunction–resembles–MolecularFunction	MFrMF	11,239	3710	3710
OligogenicCombination–involves\(\rightarrow\)Gene	OCiG	2700	1118	907
OligogenicCombination–causes\(\rightarrow\)Disease	OCcD	1173	1118	175
CellularComponent–resembles–CellularComponent	CCrCC	793	483	483

Each type of edge (i.e. metaedge) in the KG is defined uniquely by its source and target node types with the relationship name in between. Directed metaedges are indicated by an arrow on the relationship. We define abbreviations for each metaedge to simplify further notations. The table presents statistics on the number of corresponding edges, source nodes and target nodes for each metaedge, ordered by decreasing number of edges

ISSN: 1471-2105