Figure 3.
hERG classification results for all well-calibrated strategies, plotting the Observed Fuzziness (smaller values are better). The colored areas correspond to the 95% confidence intervals computed from the ten replicate runs. Note that the CCPnew strategy is independent of Aold size and will thus be the same in all four panels. The overall winning strategies are CCPpool and CCPAT, depending on the combination of Anew and Aold size, shifting from CCPpool to CCPAT when the number of compounds in Anew exceeds 10% of that of Aold.