Large scale study of multiple-molecule queries

Table 3 Mean performance of similarity methods.

Method	AUC	F1	BEDROC
MIN-RANK	0.981265 ± 0.004540	0.480749 ± 0.019308	0.915781 ± 0.012475
MAX-RANK	0.633951 ± 0.038076	0.030289 ± 0.009587	0.209204 ± 0.049180
SUM-RANK	0.840620 ± 0.032520	0.128860 ± 0.032596	0.490227 ± 0.066911
MAX-SIM	0.973312 ± 0.005642	0.484504 ± 0.022397	0.893180 ± 0.015592
MIN-SIM	0.717104 ± 0.034874	0.053943 ± 0.015827	0.284041 ± 0.058423
SUM-SIM	0.914782 ± 0.018230	0.341373 ± 0.032341	0.719190 ± 0.041269
NUMDEN-SIM	0.907632 ± 0.019609	0.327810 ± 0.037178	0.696596 ± 0.044681
BAYES	0.910909 ± 0.017386	0.149837 ± 0.033785	0.581197 ± 0.050633
BKD	0.980763 ± 0.004517	0.501197 ± 0.019569	0.890840 ± 0.017745
ETD	0.987087 ± 0.002653	0.508081 ± 0.020886	0.922371 ± 0.011330
TPD	0.986616 ± 0.002649	0.451587 ± 0.025795	0.906017 ± 0.014098
SUM-EH	0.935054 ± 0.010774	0.296279 ± 0.032160	0.699456 ± 0.037583
SUM-ET	0.974831 ± 0.005798	0.491106 ± 0.022314	0.897401 ± 0.015760
SUM-TP	0.974963 ± 0.005751	0.490653 ± 0.022311	0.897621 ± 0.015771

The mean performance of similarity methods across the 24 data sets with the ChemDB background. A confidence interval is provided with each measurement. The best performance in each column is listed in bold face, and all performances statistically indistinguishable (with a t-test yielding a p-value > 0.05) are listed in italics.

ISSN: 1758-2946