Primitives

The primitives system provides a modular approach to computation tasks. All primitives inherit from the base Primitive class and implement a standardized interface.

Primitive Base Class

The Primitive class is an abstract base class that all computation primitives inherit from.

Featurize Primitive

The Featurize primitive transforms raw molecular data into machine learning features using DeepChem’s featurizers.

Train Primitive

The Train primitive builds and trains machine learning models on featurized datasets.

Evaluate Primitive

The Evaluate primitive assesses model performance using various metrics and generates evaluation reports.

Infer Primitive

The Infer primitive runs predictions on new data using previously trained models.

TVTSplit Primitive

The TVTSplit primitive performs train-validation-test splitting on datasets using various splitter types.

DEL Denoise Primitive

The DEL Denoise primitive scores DNA-encoded library (DEL) screening data to identify compounds strongly enriched in the target selection over background noise.