. Author manuscript; available in PMC: 2015 Jun 2.

Published in final edited form as: J Biomed Inform. 2013 Apr 6;46(4):585–593. doi: 10.1016/j.jbi.2013.04.001

Table 4.

Comparison of bootstrapping approach with the dictionary-based approach with drugs and genes from PharmGKB PGx-specific drug-gene pairs as input.

Method	Cutoff (%)	N	Precision	Recall	F1
Bootstrapping (Seed “Warfarin–CYP2C9”)	50	1,311	0.354	0.297	0.323
Bootstrapping (Seed “Warfarin–CYP2C9”)	100	2,622	0.219	0.368	0.274
Bootstrapping (Seeds 68 FDA drug-gene pairs)	50	2,383	0.260	0.397	0.314
Bootstrapping (Seeds 68 FDA drug-gene pairs)	100	4,767	0.171	0.522	0.258
Dictionary-based Approach	50	6,222	0.178	0.711	0.285
Dictionary-based Approach	100	12,444	0.125	1.000	0.223
None (co-occurrence with 19,055 genes and 6,516 drugs)	50	51,458	0.024	0.783	0.046
None (co-occurrence with 19,055 genes and 6,516 drugs)	100	102,916	0.015	1.000	0.030