Rescoring allows users to calculate additional scores for molecules using predefined scoring workflows.
A rescoring workflow can combine different types of scoring components, including molecular descriptors from RDKit, QSAR models, API-based scorers, and Iktos scores such as 3D Structure-based scores.
Once configured, rescoring can be used in three main ways:
1. Standalone rescoring
Users can apply a rescoring workflow to a SMILES input or to an existing dataset. The result is a new dataset where the calculated scores are added as new columns.
2. Rescoring generated molecules
Rescoring can be applied after molecule generation (post-processing) to evaluate the generated SMILES. The calculated scores are then available in the parallel coordinates plot, making it easier to filter and prioritize compounds, especially after novelty-based or one-step-forward generations.
NOTE: scorers that include 3D Structure-based scores cannot be associated to a generator. For more details, see the corresponding page in the documentation.
3. Rescoring selected molecules after generation
Users can select any number of molecules from a generator output and apply a predefined rescoring workflow. The rescored molecules can then be saved as a Makya dataset.