Analyzing Learned Molecular Representations for Property Prediction

Benchmark Model Rank Results
molecular-property-prediction-onD-MPNN#3RMSE: 0.683
molecular-property-prediction-on-bace-1D-MPNN#10ROC-AUC: 80.9
molecular-property-prediction-on-bbbp-1D-MPNN#17ROC-AUC: 71.0
molecular-property-prediction-on-clintox-1D-MPNN#6ROC-AUC: 90.6
molecular-property-prediction-on-esolD-MPNN#19RMSE: 1.050
molecular-property-prediction-on-freesolvD-MPNN#16RMSE: 2.082
molecular-property-prediction-on-qm7D-MPNN#6MAE: 103.5
molecular-property-prediction-on-qm8D-MPNN#2MAE: 0.0190
molecular-property-prediction-on-qm9D-MPNN#2MAE: 0.00814
molecular-property-prediction-on-sider-1D-MPNN#14ROC-AUC: 57.0
molecular-property-prediction-on-tox21-1D-MPNN#9ROC-AUC: 75.9
molecular-property-prediction-on-toxcast-1D-MPNN#4ROC-AUC: 65.5