1.0

Inside ckpt.tar.gz:

• molmp_0: pre-trained model for MolMP
• molrnn_0: pre-trained model for MolRNN
• scaffold_0: model for the scaffold-based generation task
• prop_0: model for property based generation
• kinase_0: model for generating kinase inhibitors
• kinase_rf: random forest model for kinase activity prediction

Inside samples.tar.gz:

• molmp.smi: samples by MolMP
• molrnn.smi: samples by MolRNN

Inside datasets.tar.gz:

• ChEMBL.txt: training set for unconditioned model
• ChEMBL_scaffold.txt: training set labeled with scaffold fingerprint
• ChEMBL_prop.txt: training set labeled with QED and SAscore
• ChEMBL_k.txt: training set labeled with predicted kinase activity
• ChEMBL_rm.txt: training set with active compounds in the test set removed
• atom_types.txt: specification of atom types
• kinase.tsv: activity dataset for kinase
• scaffolds.smi: scaffolds extracted from drugbank