Table 5.
Statistics for Predictions Made by the Categorical Machine Learning Model for Predicting Deamidation Liability on the Independent Validation Set
| Statistic | Categorical Model |
|---|---|
| Accuracy | 93.3% |
| MCC | 0.691 |
| Precision | 81.0% |
| Recall | 65.4% |
| Specificity | 97.6% |
| Negative predictive value | 94.8% |
| Miss rate | 34.6% |
| Fallout | 19.0% |
| False discovery rate | 2.4% |
| False omission rate | 5.2% |
Notably, on the independent validation set containing non-antibody proteins our model was able to achieve 93.3% accuracy and a Matthews correlation coefficient (MCC) of 0.691.