Skip to main content

Advertisement

Table 5 Mean difference in performance of similarity methods.

From: Large scale study of multiple-molecule queries

Method AUC F1 BEDROC
MIN-RANK 0.005821 ± 0.002344 0.027332 ± 0.006678 0.006590 ± 0.003236
MAX-RANK 0.353136 ± 0.036644 0.477792 ± 0.020875 0.713166 ± 0.043907
SUM-RANK 0.146467 ± 0.030849 0.379221 ± 0.028884 0.432144 ± 0.059736
MAX-SIM 0.013775 ± 0.003534 0.023577 ± 0.005890 0.029191 ± 0.007341
MIN-SIM 0.269982 ± 0.033128 0.454138 ± 0.023192 0.638330 ± 0.052591
SUM-SIM 0.072305 ± 0.016279 0.166707 ± 0.025321 0.203181 ± 0.032996
NUMDEN-SIM 0.079455 ± 0.017748 0.180270 ± 0.031082 0.225774 ± 0.036813
BAYES 0.076178 ± 0.016079 0.358244 ± 0.032005 0.341173 ± 0.045043
BKD 0.006324 ± 0.002360 0.006884 ± 0.008400 0.031531 ± 0.008401
ETD --- --- ---
TPD 0.000471 ± 0.000443 0.056493 ± 0.010274 0.016353 ± 0.004338
SUM-EH 0.052032 ± 0.009065 0.211802 ± 0.026177 0.222915 ± 0.031105
SUM-ET 0.012256 ± 0.003648 0.016975 ± 0.006151 0.024970 ± 0.007082
SUM-TP 0.012124 ± 0.003584 0.017428 ± 0.004657 0.024749 ± 0.007079
  1. The mean difference in performance of similarity methods from that of the best method across the 24 data sets with the ChemDB background. 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.