Skip to main content
Fig. 1 | Journal of Cheminformatics

Fig. 1

From: An explainability framework for deep learning on chemical reactions exemplified by enzyme-catalysed reaction classification

Fig. 1

Explaining enzyme-catalysed reaction classifications. Enzyme-catalysed reactions (1) are classified using three different models. Model ECX predicts the class (2a), model ECXY predicts the class and subclass (2b), and model ECXYZ predicts the class, the subclass, and the sub-subclass (2c). The input to all models is a binary DRFP encoded reaction SMILES that allows a mapping between input and molecular fragment. Using Shapley additive explanations (SHAP), the inputs’ influence on the classification can be traced back to a molecular fragment (2a–c)

Back to article page