Improving Molecular Graph Neural Network Explainability with Orthonormalization and Induced Sparsity

TL;DR

This paper introduces two regularization techniques, BRO and Gini regularization, to improve the explainability and accuracy of molecular graph neural networks, making their predictions more interpretable for medicinal chemists.

Contribution

The paper proposes novel regularization methods, BRO and Gini, that enhance GCNN explainability and accuracy, with potential applicability to other neural network types.

Findings

Regularization improves GCNN attribution accuracy on benchmarks.

Chemists prefer explanations from regularized models.

Methods are applicable beyond GCNNs.

Abstract

Rationalizing which parts of a molecule drive the predictions of a molecular graph convolutional neural network (GCNN) can be difficult. To help, we propose two simple regularization techniques to apply during the training of GCNNs: Batch Representation Orthonormalization (BRO) and Gini regularization. BRO, inspired by molecular orbital theory, encourages graph convolution operations to generate orthonormal node embeddings. Gini regularization is applied to the weights of the output layer and constrains the number of dimensions the model can use to make predictions. We show that Gini and BRO regularization can improve the accuracy of state-of-the-art GCNN attribution methods on artificial benchmark datasets. In a real-world setting, we demonstrate that medicinal chemists significantly prefer explanations extracted from regularized models. While we only study these regularizers in the context of GCNNs, both can be applied to other types of neural networks