Fig. 6From: In silico prediction of UGT-mediated metabolism in drug-like molecules via graph neural networkA The performance of top-k accuracy in different iterations of message passing (n_layers = 1, 2, 3, 4, 5); B different batch size (10, 16, 20, 32, 64) and learning rate (0.0001, 0.0003, 0.001, 0.003) were employed to train the SOM prediciton model; C the top-1, top-2 and top-3 accuracy was used to assess and compare all of these modelsBack to article page