Skip to main content
. Author manuscript; available in PMC: 2023 Feb 9.
Published in final edited form as: Comput Toxicol. 2018 Nov;8:1–12. doi: 10.1016/j.comtox.2018.08.003

Table 4.

Mean performance data of every variation of four ER assays for the two ER models and how they compare to using the burst flag hit-call data or the presence of a structural alert in isolation. Values in parentheses are 95% confidence intervals (no confidence interval could be calculated for the presence of a structural alert as there were no parameters to change). MCC – Matthews Correlation Coefficient.

Active call in burst hit-call matrix Presence of structural alert only Active call in burst hit-call matrix AND presence of structural alert Active call in burst hit-call matrix OR presence of structural alert
Sensitivity 67% (48–80%) 69% 52% (40–62%) 84% (77–89%)
Specificity 92% (85–97%) 73% 98% (96–99%) 67% (61–71%)
Accuracy 90% (84–94%) 72% 94% (93–95%) 68% (63–71%)
MCC 0.466 (0.38–0.575) 0.234 0.546 (0.484–0.617) 0.27 (0.244–0.292)