Fig. 5

Predictive performances of our machine-learning approaches on the SMD sets relative to MolNet and the respective original models. With the exception of MUV, the metric used is ROC-AUC. The higher the y-axis is, the better the model performs
Predictive performances of our machine-learning approaches on the SMD sets relative to MolNet and the respective original models. With the exception of MUV, the metric used is ROC-AUC. The higher the y-axis is, the better the model performs