It would be interesting to add a feature for model explainability like GNNExplainer, PGExplainer, or any other explainability method.