Joint embedding–classifier learning for interpretable collaborative filtering

BMC Bioinformatics

Table 11 Drug recommendations in decreasing order of recommendation score (rounded to the closest third decimal place) for melanoma after running JELI on the TRANSCRIPT data set

DrugBank ID	Drug name	Score	Label	Drug family	ATC
DB00928	Azacitidine	0.927	0	Pyrimidine nucleoside analogue	L01BC07
DB00331	Metformin	0.926	0	Biguanide antihyperglycemic	A10BA02
DB01067	Glipizide	0.924	0	Sulfonylurea medication	A10BB07
DB00685	Trovafloxacin	0.924	0	Broad spectrum antibiotic	unknown
DB01197	Captopril	0.924	0	ACE inhibitor	C09AA01
DB00983	Formoterol	0.924	0	Long-acting beta2-adrenergic	R03CC15
				receptor agonist
DB00526	Oxaliplatin	0.921	0	Platinum-based chemotherapy	L01XA03
				agent
DB01064	Isoprenaline	0.920	0	Catecholamine non-selective	C01CA02
				beta-adrenergic agonist
DB00281	Lidocaine	0.704	1	Tertiary amine class Ib	C05AD01
				antiarrhythmic agent

The identifiers and drug families come from DrugBank [63], whereas the Anatomical Therapeutic Code (ATC) is extracted from PubChem [64]. The ATC classifies drugs into hierarchical classes depending on their mechanisms of action. The label (0: unknown, 1: positive) in the TRANSCRIPT data set corresponds to the initial drug-disease class for melanoma used for the training of JELI

ISSN: 1471-2105