. Author manuscript; available in PMC: 2010 Dec 1.

Published in final edited form as: J Biomed Inform. 2009 Jun 16;42(6):990–1003. doi: 10.1016/j.jbi.2009.05.010

Table 18.

This table shows the range of LOEs used by 1,152 belief criteria strategies that had perfect sensitivity and positive predictive value and optimized the DIKB’s coverage and agreement with the validation set. The LOEs that were chosen as belief criteria for supporting and refuting evidence were always the same for the assertions shown in the table. The columns Evidence for and Evidence against show the distribution of evidence items across the LOEs. Not shown are the assertion types for which varying the LOEs chosen as belief criteria would have no effect on the DIKB’s prediction performance (Section 2.3.4).

Assertion type	Belief criteria	Evidence for	Evidence against
inhibits	I: LOE-1, LOE-2	n=11 ∼ LOE-1(91%), LOE-2(9%)	n=4 ∼ LOE-1(50%), LOE-2(50%)
substrate-of	D: LOE-3	n=29 ∼ LOE-1(10%), LOE-2(21%) LOE-3(17%), LOE-4(21%) LOE-5(31%)	n=11 ∼ LOE-1(0%), LOE-2(0%) LOE-3(36%), LOE-4(9%) LOE-5(55%)
is-not-substrate-of	D: LOE-1 → LOE-3	n=11 ∼ LOE-1(0%), LOE-2(0%) LOE-3(36%), LOE-4(9%) LOE-5(55%)
primary-total-clearance-mechanism	C: LOE-1 → LOE-3	n=14 ∼ LOE-1(21%), LOE-2(64%), LOE-3(14%)
primary-total-clearance-enzyme	E: LOE-3	n=8 ∼ LOE-1(25%), LOE-2(62%) LOE-3(13%)
primary-metabolic-clearance-enzyme	F: LOE-1 → LOE-4	n=3 ∼ LOE-1(0%), LOE-2(0%) LOE-3(67%), LOE-4(33%)	n=1 ∼ LOE-1(100%), LOE-2(0%) LOE-3(0%), LOE-4(0%)
controls-formation-of ⁶	D: LOE-1 → LOE-3, LOE-4, LOE-5	n=17 ∼ LOE-1(0%), LOE-2(0%), LOE-3(12%), LOE-4(6%), LOE-5(82%)
has-metabolite⁶	G: LOE-1 → LOE-2 LOE-3, LOE-4	n=27 ∼ LOE-1(19%), LOE-2(0%) LOE-3(44%), LOE-4(37%)

⁶

No belief criteria strategy in the set of 1,152 used LOE-4 or any lower-ranking LOE for the controls-formation-of assertion type and LOE-3 or any lower-ranking LOE for the has-metabolite assertion type.