Table 6. Gal4 monomers predict one another.
Feature vector input attributes based on in silico mutagenesis using tessellation of the Gal4 dimer. The full data set was split into two subsets of 1,068 Gal4 variants each by monomeric chain. Variants from one chain were predicted using a model trained with the variants from the other chain.
| Method | Se | Sp | PPV | BAR | MCC | AUC |
|---|---|---|---|---|---|---|
| A chain—training/B chain—testing: | ||||||
| RF | 0.96 | 0.71 | 0.82 | 0.83 | 0.71 | 0.96 |
| SVR (r = 0.60) | 0.95 | 0.54 | 0.75 | 0.75 | 0.56 | – |
| B chain—training/A chain—testing: | ||||||
| RF | 0.96 | 0.85 | 0.90 | 0.90 | 0.82 | 0.98 |
| SVR (r = 0.66) | 0.94 | 0.56 | 0.75 | 0.75 | 0.56 | – |