Table 14.

Parameters of the selected generalized weighted-kNN models as implemented in OPERA.

End point	Number of chemicals from the prediction set	Descriptors	Training (5-fold cross-validation) $\text{BA}/Q^2$*	Test set $\text{BA}/R^2$*
VT	23,767	21	0.79	0.77
NT	30,971	11	0.90	0.89
EPA categories	25,487	15	0.79	0.81
GHS categories	25,720	15	0.78	0.79
$\mathrm{L}{\mathrm{D}}_{50}$ values	28,954	23	0.79	0.81

^*Balanced accuracy (BA) values are reported for performance metrics, except for the $\mathrm{L}{\mathrm{D}}_{50}$ value end point for which the coefficient of determination $R^2$ and its equivalent for cross-validation $Q^2$ are reported for the training and test sets, respectively. Note: EPA, U.S. Environmental Protection Agency; GHS, U.N. Globally Harmonized System of Classification and Labeling of Chemicals; $\mathrm{L}{\mathrm{D}}_{50}$_, dose of a substance that would be expected to kill half the animals in a test group; NT, nontoxic/toxic; VT, very toxic/not very toxic. The number of prediction set chemicals represents the total for that data set, not the number that fell beneath the threshold (e.g., for VT).