A molecular scaffold is the core structural framework shared by related molecules. Compounds with the same scaffold often form a chemical series, making the scaffold a useful unit for organizing chemical structure and design. section 7.2.2
Scaffolds support several practical drug-discovery workflows:
- Scaffold splitting keeps structurally similar compounds in the same dataset split, making model evaluation harder and reducing optimistic validation caused by related molecules crossing train and test sets. section 4.1.1
- Scaffold decoration holds the core constant while varying attached R-groups to create analogs and investigate how structural changes affect activity or other properties. section 7.2.2
- Scaffold hopping replaces the entire core with an alternative scaffold while preserving R-groups, seeking similar biological activity without inherited toxicity or patentability limitations. section 7.2.2
Choosing an alternative scaffold
Replacing a benzene ring with a pyridine ring can preserve a similar structure, while a pyrimidine scaffold may fail if its two nitrogens cannot support the required R-group orientation.
Listing 4.1 Load AqSolDB with Scaffold-based Splits
from tdc.single_pred import ADME data = ADME(name = 'Solubility_AqSolDB') split = data.get_split(method = 'scaffold') train_df, valid_df, test_df = split['train'], split['valid'], split['test']
A scaffold-based split can be created with TDC's Solubility_AqSolDB interface and separated into training, validation, and test data. section 4.1.1
