Table 1.

Performance comparison of pharmacogenomic (PGx) relationships obtained from our proposed pipeline with other benchmark datasets (OMIM, CTD, and PharmGKB).

Context type	TP	TN	FP	FN	Sensitivity	Specificity	Efficacy	Precision	Recall	F measure	Accuracy
Our pipeline with “PharmGKB”	1,509	208	254	78	82.6	86.9	88.2	0.904	0.930	0.923	89.1
Our pipeline with “OMIM”	2,225	–	79	–	78.0	77.5	81.8	0.600	0.681	0.764	59.3
Our pipeline with “CTD”	1,776	153	375	–	70.7	65.5	72.2	0.729	0.803	0.801	79.7
Our pipeline with (“PharmGKB” AND “OMIM” AND “CTD”)	1,875	102	275	75	82.3	84.4	93.3	0.896	0.852	0.828	94.7

PharmGKB corpus compared to that of our pipeline and the articles extracted in these datasets. The formulae used for calculating the accuracy of the proposed pipeline compared to the other datasets (ref. Supplementary Table 5 for detailed analysis).