Skip to main content

Advertisement

Table 1 Comparison of the performance of the different methods

From: An exploration strategy improves the diversity of de novo ligands using deep reinforcement learning: a case for the adenosine A2A receptor

  DrugEx (Pre-trained) DrugEx (Fine-tuned) REINVENT ORGANIC Pre-trained Fine-tuned
ε 0.01 0.01 0.1 0.1 0.01 0.01 0.1 0.1
β 0.0 0.1 0.0 0.1 0.0 0.1 0.0 0.1
Valid SMILES 98.3% 98.9% 95.9% 98.8% 99.1% 99.0% 98.2% 97.5% 98.8% 99.8% 93.9% 96.2%
Desired SMILES 97.5% 98.0% 74.6% 80.9% 98.3% 98.5% 94.4% 94.5% 98.2% 99.8% 0.7% 47.9%
Unique SMILES 96.5% 96.3% 73.0% 80.0% 96.5% 96.6% 84.8% 86.0% 95.8% 94.8% 0.7% 22.7%
Diversity 0.74 0.75 0.80 0.80 0.75 0.74 0.80 0.80 0.75 0.67 0.83 0.82
  1. These methods included DrugEx with different ε, β and Gφ (shown in the parentheses), REINVENT, ORGANIC, the pre-trained network, and the fine-tuned network (both without using DrugEx)