TABLE 4.
Fragment similarity (Frag), Scaffold similarity (Scaff).
| Model | Frag (↑) | Scaf (↑) | ||
|---|---|---|---|---|
| Test | TestSF | Test | TestSF | |
| Train | 1.0 | 0.999 | 0.991 | 0.0 |
| HMM | 0.5754 ± 0.1224 | 0.5681 ± 0.1218 | 0.2065 ± 0.0481 | 0.049 ± 0.018 |
| NGram | 0.9846 ± 0.0012 | 0.9815 ± 0.0012 | 0.5302 ± 0.0163 | 0.0977 ± 0.0142 |
| Combinatorial | 0.9912 ± 0.0004 | 0.9904 ± 0.0003 | 0.4445 ± 0.0056 | 0.0865 ± 0.0027 |
| CharRNN | 1.0 ± 0.0 | 0.998 ± 0.0 | 0.924 ± 0.006 | 0.11 ± 0.008 |
| VAE | 0.999 ± 0.0 | 0.998 ± 0.0 | 0.939 ± 0.002 | 0.059 ± 0.01 |
| AAE | 0.991 ± 0.005 | 0.99 ± 0.004 | 0.902 ± 0.037 | 0.079 ± 0.009 |
| JTN-VAE | 0.9965 ± 0.0003 | 0.9947 ± 0.0002 | 0.8964 ± 0.0039 | 0.1009 ± 0.0105 |
| LatentGAN | 0.999 ± 0.003 | 0.998 ± 0.003 | 0.886 ± 0.015 | 0.1 ± 0.006 |
Reported (mean ± SD) over three independent model initializations. Results for random test set (Test) and scaffold split test set (TestSF).