Table 6

Mean difference in performance of similarity methods across MUV data sets.

Method

AUC

F1

BEDROC

MIN-RANK

0.054811 ± 0.007343

0.007274 ± 0.009481

0.011562 ± 0.013673

MAX-RANK

0.276475 ± 0.021613

0.139500 ± 0.023056

0.295164 ± 0.036878

SUM-RANK

0.187161 ± 0.021317

0.135635 ± 0.020176

0.251934 ± 0.029684

MAX-SIM

0.071096 ± 0.008692

0.000284 ± 0.005657

0.044583 ± 0.013589

MIN-SIM

0.252742 ± 0.020037

0.136318 ± 0.021634

0.286359 ± 0.031277

SUM-SIM

0.168871 ± 0.018754

0.104246 ± 0.014075

0.203296 ± 0.017032

NUMDEN-SIM

0.141477 ± 0.019134

0.096007 ± 0.014050

0.179707 ± 0.018725

BAYES

0.143037 ± 0.016147

0.115277 ± 0.014736

0.180010 ± 0.012881

BKD

0.001826 ± 0.001822

0.011572 ± 0.009888

0.002483 ± 0.005203

ETD

---

0.015242 ± 0.010510

---

TPD

0.010170 ± 0.003688

0.004709 ± 0.007578

0.003758 ± 0.007367

SUM-EH

0.106459 ± 0.013059

0.056296 ± 0.007868

0.126204 ± 0.012854

SUM-ET

0.052051 ± 0.010111

0.001559 ± 0.004737

0.032111 ± 0.012709

SUM-TP

0.056095 ± 0.009880

---

0.033664 ± 0.012780

The mean difference in performance of similarity methods from that of the best method across the 17 MUV data sets. A confidence interval is provided with each measurement. All performances statistically indistinguishable (with a paired t-test yielding a p-value > 0.05) are listed in italics.

Nasr et al. Journal of Cheminformatics 2009 1:7   doi:10.1186/1758-2946-1-7