Table 6. Composition of Ternary Training and Test Data Sets.
training set1 | test set2 | |
---|---|---|
nonsensitizer | 510 | 124 |
weak to moderate sensitizer | 279 | 66 |
strong to extreme sensitizer | 65 | 19 |
total no. compounds | 854 | 209 |
Compared to the binary training set, 173 compounds have been removed because of missing multiclass labels and one compound has been rejected because of conflicting ternary class labels.
Compared to the binary test set, 47 compounds have been removed because of missing multiclass labels and one compound has been rejected because of conflicting ternary class labels.