Table 2.
Predictivity of the LLNA for DASS DB chemicals which have LLNA and HDSG data available (n = 56). Three subsets have been calculated: all (N = 56), LogP<3.5 (N = 44), and LogP≥3.5 (N = 12).
| Set | FN | FP | TN | TP | N | Acc. | Bal-Acc | Sens. | Spec. | FDR 2) | Prev. |
|---|---|---|---|---|---|---|---|---|---|---|---|
| All | 3 | 7 | 2 | 44 | 56 | 0.82 | 0.58 | 0.94 | 0.221) | 0.14 | 0.8 |
| LogP<3.5 | 3 | 2 | 1 | 38 | 44 | 0.89 | 0.63 | 0.93 | 0.33 | 0.05 | 0.93 |
| LogP≥3.5 | 0 | 5 | 1 | 6 | 12 | 0.58 | 0.58 | 1 | 0.17 | 0.46 | 0.5 |
1
The specificity values are based on a relatively low number of chemicals compared to sensitivity values. Clopper Pearson 95% confidence intervals were thus calculated: The CI for the specificity of 0.22 is [0.02, 0.60].
2
False-discovery rate FDR indicates the proportion of false positive within the positive predictions (FDR=FP/(FP + TP)).