The models in Cerella estimate the uncertainty in every individual prediction. This is one of the big advantages of the Alchemite method underlying Cerella.

In some cases, the uncertainty in a prediction is so large that it cannot distinguish confidently between the highest and lowest observed value for the property in the training data set. You can consider this as the compound lying far outside of the domain of applicability of the model. In this case, Cerella will not return a predicted value.

More resources