Molecular Property Prediction on BBBP (test)
0.92ROC-AUCProtoW-L2
Evaluation Results
| Method | Links | |
|---|---|---|
| ProtoW-L2Contrastive regularization=false, Distance metric=L22020.06 | 0.92 | |
| ProtoW-L2Contrastive regularization=true, Distance metric=L22020.06 | 0.92 | |
| ProtoW-DotContrastive regularization=false, Distance metric=Dot2020.06 | 0.919 | |
| ProtoW-DotContrastive regularization=true, Distance metric=Dot2020.06 | 0.919 | |
| ProtoS-L2Contrastive regularization=true, Distance metric=L22020.06 | 0.918 | |
| D-MPNNPooling method=summation2020.06 | 0.915 | |
| D-MPNN+TopK PoolPooling method=TopK Pool2020.06 | 0.912 | |
| Fingerprint+MLP2020.06 | 0.911 | |
| D-MPNN+SAG PoolPooling method=SAG Pool2020.06 | 0.901 | |
| GINPooling method=summation2020.06 | 0.9 | |
| MolSight S6prediction_mode=Yes/No parsed prediction2026.05 | 0.893 | |
| GATPooling method=summation2020.06 | 0.888 | |
| GIT-MolInput Modality=Graph + SMILES2023.08 | 0.739 | |
| GraphLoG2025.04 | 0.725 | |
| GPF-plusPre-training Strategy=GCL2022.09 | 0.7218 | |
| Mole-BERT2023.08 | 0.719 | |
| GIT-MolInput Modality=SMILES2023.08 | 0.719 | |
| MoleBERTBackbone=MoleBERT, Training=Pre-trained2025.04 | 0.719 | |
| GPFPre-training Strategy=GCL2022.09 | 0.7111 | |
| GIT-MolInput Modality=Graph2023.08 | 0.711 | |
| GraphMVP2023.08 | 0.708 | |
| MGSSL (DFS)Backbone=GIN, Evaluation Protocol=Fine-tuning2021.10 | 0.705 | |
| KV-PLM2023.08 | 0.705 | |
| MoMu2023.08 | 0.705 | |
| GraphMAE+GFSEBackbone=GraphMAE, SE=GFSE2025.04 | 0.705 | |
| SUPThardPre-training Strategy=ContextPred2024.02 | 0.7018 | |
| SUPTsoftPre-training Strategy=ContextPred2024.02 | 0.7006 | |
| SUPTsoftPre-training Strategy=EdgePred2024.02 | 0.6993 | |
| SUPTsoftPre-training Strategy=GCL2024.02 | 0.6979 | |
| MGSSL (BFS)Backbone=GIN, Evaluation Protocol=Fine-tuning2021.10 | 0.697 | |
| SUPThardPre-training Strategy=GCL2024.02 | 0.6968 | |
| FTPre-training Strategy=ContextPred2022.09 | 0.6965 | |
| FTPre-training Strategy=ContextPred2024.02 | 0.6965 | |
| GPFPre-training Strategy=EdgePred2022.09 | 0.6957 | |
| SUPThardPre-training Strategy=EdgePred2024.02 | 0.6951 | |
| GraphCL2025.04 | 0.695 | |
| GraphMAEBackbone=GraphMAE, Training=Pre-trained2025.04 | 0.695 | |
| FTPre-training Strategy=GCL2022.09 | 0.6949 | |
| FTPre-training Strategy=GCL2024.02 | 0.6949 | |
| GPPT (w/o ol)Pre-training Strategy=EdgePred2022.09 | 0.6943 | |
| GPPTw/o olPre-training Strategy=EdgePred2024.02 | 0.6943 | |
| GPFPre-training Strategy=EdgePred2024.02 | 0.6936 | |
| GPF-plusPre-training Strategy=EdgePred2024.02 | 0.6931 | |
| GraphPromptPre-training Strategy=EdgePred2022.09 | 0.6929 | |
| GraphPromptPre-training Strategy=EdgePred2024.02 | 0.6929 | |
| GPF-plusPre-training Strategy=ContextPred2022.09 | 0.6915 | |
| GINE+GFSEBackbone=GINE, SE=GFSE2025.04 | 0.691 | |
| GPF-plusPre-training Strategy=EdgePred2022.09 | 0.6906 | |
| GPFPre-training Strategy=ContextPred2024.02 | 0.6904 | |
| MoleBERT+GFSEBackbone=MoleBERT, SE=GFSE2025.04 | 0.689 | |
| GPF-plusPre-training Strategy=ContextPred2024.02 | 0.6859 | |
| GPFPre-training Strategy=ContextPred2022.09 | 0.6848 | |
| SUPThardPre-training Strategy=AttrMasking2024.02 | 0.6839 | |
| SUPTsoftPre-training Strategy=AttrMasking2024.02 | 0.6837 | |
| GPFPre-training Strategy=GCL2024.02 | 0.6811 | |
| InfomaxBackbone=GIN, Evaluation Protocol=Fine-tuning2021.10 | 0.681 | |
| GPFPre-training Strategy=AttrMasking2022.09 | 0.6809 | |
| GroverBackbone=GIN, Evaluation Protocol=Fine-tuning2021.10 | 0.68 | |
| GPS+GFSEBackbone=GPS, SE=GFSE2025.04 | 0.68 | |
| SSP2025.04 | 0.679 | |
| GPS+RWSEBackbone=GPS, SE=RWSE2025.04 | 0.679 | |
| GPS+LapPEBackbone=GPS, SE=LapPE2025.04 | 0.679 | |
| GraphCL2023.08 | 0.678 | |
| GPS+GPSEBackbone=GPS, SE=GPSE2025.04 | 0.678 | |
| GPF-plusPre-training Strategy=GCL2024.02 | 0.6777 | |
| GPF-plusPre-training Strategy=AttrMasking2022.09 | 0.6771 | |
| GINEBackbone=GINE, Training=from-scratch2025.04 | 0.677 | |
| SUPTsoftPre-training Strategy=Infomax2024.02 | 0.6763 | |
| FTPre-training Strategy=Infomax2022.09 | 0.6755 | |
| FTPre-training Strategy=Infomax2024.02 | 0.6755 | |
| GPT-GNNBackbone=GIN, Evaluation Protocol=Fine-tuning2021.10 | 0.675 | |
| GPFPre-training Strategy=AttrMasking2024.02 | 0.6749 | |
| GPF-plusPre-training Strategy=AttrMasking2024.02 | 0.6732 | |
| SUPThardPre-training Strategy=Infomax2024.02 | 0.6729 | |
| GPF-plusPre-training Strategy=Infomax2022.09 | 0.6717 | |
| GPSBackbone=GPS, Training=from-scratch2025.04 | 0.671 | |
| GPF-plusPre-training Strategy=Infomax2024.02 | 0.6703 | |
| GCCBackbone=GIN, Evaluation Protocol=Fine-tuning2021.10 | 0.669 | |
| GPFPre-training Strategy=Infomax2022.09 | 0.6683 | |
| GINE+RWSEBackbone=GINE, SE=RWSE2025.04 | 0.667 | |
| GPFPre-training Strategy=Infomax2024.02 | 0.6658 | |
| FTPre-training Strategy=EdgePred2022.09 | 0.6656 | |
| FTPre-training Strategy=EdgePred2024.02 | 0.6656 | |
| MoleBERT+RWSEBackbone=MoleBERT, SE=RWSE2025.04 | 0.665 | |
| GINE+GPSEBackbone=GINE, SE=GPSE2025.04 | 0.664 | |
| GraphMAE+RWSEBackbone=GraphMAE, SE=RWSE2025.04 | 0.664 | |
| FTPre-training Strategy=AttrMasking2022.09 | 0.6633 | |
| FTPre-training Strategy=AttrMasking2024.02 | 0.6633 | |
| GINE+LapPEBackbone=GINE, SE=LapPE2025.04 | 0.658 | |
| Attribute maskingBackbone=GIN, Evaluation Protocol=Fine-tuning2021.10 | 0.656 | |
| No pretrainBackbone=GIN, Evaluation Protocol=Fine-tuning2021.10 | 0.655 | |
| GPPTPre-training Strategy=EdgePred2022.09 | 0.6413 | |
| GPPTPre-training Strategy=EdgePred2024.02 | 0.6413 | |
| Qwen2-VL-2Bprediction_mode=Yes/No parsed prediction2026.05 | 0.475 |