Table 3.
Average AUC values using the prediction models with different numbers of basis compounds
Models | Number of basis compounds | ||||||
---|---|---|---|---|---|---|---|
20 | 40 | 60 | 80 | 100 | 120 | 140 | |
NCI antiviral dataset | |||||||
AP | 70.7 | 72.4 | 73.3 | 73.9 | 74.0 | 72.9 | 72.9 |
MCS c2 | 73.0 | 74.6 | 74.6 | 75.8 | 75.5 | 75.4 | 75.2 |
Hybrid c2 | 74.4 | 75.2 | 75.4 | 76.2 | 76.1 | 75.4 | 75.6 |
NCI anticancer dataset | |||||||
AP | 69.5 | 72.4 | 72.6 | 73.2 | 73.9 | 74.7 | 73.7 |
MCS c2 | 71.0 | 74.2 | 75.2 | 75.5 | 75.9 | 76.6 | 76.4 |
Hybrid c2 | 74.4 | 75.9 | 75.9 | 76.1 | 76.5 | 77.2 | 76.9 |
For the NCI antiviral dataset, 25 000 randomly selected compounds were used as the training set. For the NCI anticancer dataset, 5000 randomly selected compounds were used as the training set.