Molecular Property Prediction
Property prediction on molecular graphs (QM9, MoleculeNet), including pretrained DimeNet and SchNet.
-
Pre-trained DimeNet models on QM9
Evaluate the official pre-trained DimeNet (Gasteiger et al., 2020) or DimeNet++ models on QM9: one model per target property.
QM9·DimeNet · DimeNetPlusPlus -
Pre-trained SchNet models on QM9
Evaluate the official pre-trained SchNet models (Schütt et al., 2017) on QM9: one model per target property.
QM9·SchNet -
Predicting quantum properties of molecules with an MPNN (QM9)
Predict a quantum-mechanical property (here the dipole moment, target 0) of small molecules from their structure.
QM9·NNConv · Set2Set -
Predicting solubility with AttentiveFP (ESOL)
Predict how well molecules dissolve in water.
ESOL (MoleculeNet)·AttentiveFP -
Predicting solubility with Principal Neighbourhood Aggregation
Predict how well molecules dissolve in water.
ESOL (MoleculeNet)·BatchNorm · PNAConv -
Predicting solubility with a graph transformer (GraphGPS)
Predict how well molecules dissolve in water.
ESOL (MoleculeNet)·GINEConv · GPSConv