Skip to main content

Advertisement

Table 3 Dependency on the χ cutoff value using models generated from datasets with 250 actives (n) on 10,000 compounds in total (N)

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

Metric χ λ
0.5% 1% 2.5% 5% 10%
PM 0.52 ± 0.25 0.57 ± 0.15 0.60 ± 0.08 0.60 ± 0.06 0.60 ± 0.04 1
ROCE 1.60 ± 1.19 1.59 ± 0.81 1.58 ± 0.51 1.55 ± 0.35 1.52 ± 0.23
EF 1.54 ± 1.10 1.56 ± 0.76 1.55 ± 0.48 1.53 ± 0.33 1.50 ± 0.22
REF 3.86 ± 2.75 3.89 ± 1.90 3.88 ± 1.20 7.63 ± 1.65 14.97 ± 2.22
CCR 0.50 ± 0.00 0.50 ± 0.00 0.51 ± 0.01 0.51 ± 0.01 0.53 ± 0.01
MCC 0.01 ± 0.01 0.01 ± 0.01 0.01 ± 0.01 0.02 ± 0.01 0.03 ± 0.01
CKC 0.00 ± 0.01 0.01 ± 0.01 0.01 ± 0.01 0.02 ± 0.01 0.02 ± 0.01
SEN 0.01 ± 0.01 0.02 ± 0.01 0.04 ± 0.01 0.08 ± 0.02 0.15 ± 0.02
SPE 1.00 ± 0.00 0.99 ± 0.00 0.98 ± 0.00 0.95 ± 0.00 0.90 ± 0.00
PRE 0.04 ± 0.03 0.04 ± 0.02 0.04 ± 0.01 0.04 ± 0.01 0.04 ± 0.01
ACC 0.97 ± 0.00 0.97 ± 0.00 0.95 ± 0.00 0.93 ± 0.00 0.88 ± 0.00
PM 0.96 ± 0.01 0.96 ± 0.01 0.96 ± 0.00 0.94 ± 0.00 0.91 ± 0.00 20
ROCE 28.80 ± 8.24 26.73 ± 5.20 22.13 ± 2.46 16.72 ± 1.11 10.49 ± 0.34
EF 16.67 ± 2.70 16.12 ± 1.85 14.44 ± 1.03 11.99 ± 0.56 8.48 ± 0.22
REF 41.68 ± 6.74 40.30 ± 4.62 36.09 ± 2.56 59.97 ± 2.79 84.78 ± 2.18
CCR 0.54 ± 0.01 0.58 ± 0.01 0.67 ± 0.01 0.78 ± 0.01 0.88 ± 0.01
MCC 0.18 ± 0.03 0.24 ± 0.03 0.34 ± 0.03 0.40 ± 0.02 0.40 ± 0.01
CKC 0.13 ± 0.02 0.22 ± 0.03 0.34 ± 0.03 0.38 ± 0.02 0.31 ± 0.01
SEN 0.08 ± 0.01 0.16 ± 0.02 0.36 ± 0.03 0.60 ± 0.03 0.85 ± 0.02
SPE 1.00 ± 0.00 0.99 ± 0.00 0.98 ± 0.00 0.96 ± 0.00 0.92 ± 0.00
PRE 0.42 ± 0.07 0.40 ± 0.05 0.36 ± 0.03 0.30 ± 0.01 0.21 ± 0.01
ACC 0.97 ± 0.00 0.97 ± 0.00 0.97 ± 0.00 0.96 ± 0.00 0.92 ± 0.00
  1. The PM is not so much dependent on the applied cutoff value. For good models the EF and ROCE metrics decrease when the cutoff is increased, while the REF, CCR, MCC and CKC values always increase when the cutoff is increased from 2.5% up to 10%