Skip to main content

Advertisement

Fig. 1 | Journal of Cheminformatics

Fig. 1

From: The power metric: a new statistically robust enrichment-type metric for virtual screening applications with early recovery capability

Fig. 1

Illustration of the relation between cutoff χ and number of predicted actives/non-actives. Assuming a list of compounds ranked according to their predicted activity values, all compounds that are located on the left side of χ on this ranked list are predicted to be active, while all compounds that are located on the right-hand side of χ on this list are predicted to be non-active. All compounds that fall along the left-hand side of χ define the ‘selection set’; in this example this includes five compounds. The total number of compounds in the selection set is N s (here: 5), while the total number of compounds in the entire collection is N (here: 15). The number of true actives in the selection set is n s (here: 3) and the number of true actives in the entire data collection is n (here: 4). Using these abbreviations, one can define the number of true positives TP as being equal to n s , the number of true negatives TN equal to (N − N s  − n + n s ), the number of false positives FP equal to (N s  − n s ), and the number of false negatives FN being equal to (n − n s )

Back to article page