Skip to main content

Table 1 Validity, uniqueness and novelty (mean ± std) of SMILES generated after training

From: De novo design and bioactivity prediction of SARS-CoV-2 main protease inhibitors using recurrent neural network-based transfer learning

Temperature Validity (%) Uniqueness (%) Novelty (%)
0.20 100.00 ± 0.00 39.79 ± 0.27 33.21 ± 0.59
0.50 99.98 ± 0.03 99.05 ± 0.30 78.44 ± 0.78
0.60 99.95 ± 0.04 99.05 ± 0.18 81.80 ± 1.19
0.70 99.80 ± 0.10 99.58 ± 0.16 85.10 ± 0.58
0.75 99.72 ± 0.15 99.58 ± 0.12 85.85 ± 0.68
0.80 99.44 ± 0.21 99.36 ± 0.20 87.11 ± 0.59
1.00 97.21 ± 0.39 97.15 ± 0.15 88.66 ± 0.95
1.20 89.95 ± 0.23 89.84 ± 0.24 85.38 ± 0.87
  1. We sampled 2000 SMILES for each temperature in five independent runs (10,000 in total)