Strategies for Pre-training Graph Neural Networks

Benchmark Model Rank Results
molecular-property-prediction-on-qm8PretrainGNN#3MAE: 0.0200
molecular-property-prediction-on-qm9PretrainGNN#3MAE: 0.00922
molecular-property-prediction-on-toxcastPretrainGNN#3ROC-AUC: 65.7
drug-discovery-on-bbbpContextPred#4AUC: 0.687
drug-discovery-on-clintoxContextPred#4AUC: 0.726
drug-discovery-on-hiv-datasetContextPred#4AUC: 0.799
drug-discovery-on-muvContextPred#4AUC: 0.813
drug-discovery-on-siderContextPred#4AUC: 0.627
molecular-property-prediction-on-bacePretrainGNN#4ROC-AUC: 84.5
drug-discovery-on-baceContextPred#5AUC: 0.845
drug-discovery-on-toxcastContextPred#5AUC: 0.657
molecular-property-prediction-on-lipophilicityPretrainGNN#5RMSE: 0.739
molecular-property-prediction-on-tox21PretrainGNN#6ROC-AUC: 78.1
drug-discovery-on-tox21ContextPred#7AUC: 0.781
molecular-property-prediction-on-qm7PretrainGNN#7MAE: 113.2
molecular-property-prediction-on-siderPretrainGNN#12ROC-AUC: 62.7
molecular-property-prediction-on-clintoxPretrainGNN#15ROC-AUC: 72.6
molecular-property-prediction-on-freesolvPretrainGNN#20RMSE: 2.764
molecular-property-prediction-on-bbbpPretrainGNN#22ROC-AUC: 68.7