Skip to main content

Table 2 Mean property of all generated molecules for 200 validation compounds

From: MERMAID: an open source automated hit-to-lead method based on deep reinforcement learning

  Validity Uniqueness Novelty
PLogP-Single 0.298 0.981 1.0
PLogP-Multi 0.318 0.991 1.0
QED-Single 0.626 0.945 0.999
QED-Multi 0.770 0.958 0.999
  1. Validity, uniqueness and novelty of all generated molecules are shown for four models. Validity is the ratio of valid SMILES to all generated SMILES. Uniqueness is the ratio of non-duplicate molecules to all valid molecules. Novelty is the ratio of molecules that are not included in the training dataset to all valid molecules