Skip to main content

Table 4 Comparison of the percentage of generated desirable molecules from HierG2G, Seq2Seq and Transformer that are MMPs with starting molecules

From: Molecular optimization by capturing chemist’s intuition using deep neural networks

  HierG2G Seq2Seq Transformer
  Desirable (%) MMP_ 0.33_D (%) (%) MMP_ 0.50_D (%) Desirable (%) MMP_ 0.33_D (%) MMP_ 0.50_D (%) Desirable (%) MMP_ 0.33_D (%) MMP_ 0.50_D (%)
Test-Original 38.66 62.94 70.77 46.94 79.66 88.11 56.14 92.67 97.64
Test-Molecule 37.62 56.81 65.44 45.22 74.81 84.26 56.74 91.41 96.98
Test-Property 33.58 55.70 63.28 37.67 74.32 83.43 41.75 91.06 96.83
  1. Desirable represents the percentage of molecules with desirable properties among all the generated molecules. Among all these generated molecules with desirable properties for each test set, MMP_0.33_D and MMP_0.50_D represent the percentage of generated molecules that are MMPs with their corresponding starting molecules and the ratio of heavy atoms in R group to the generated molecule \(R_{group}\leq0.33\) and \(R_{group}\leq0.50\) respectively
  2. The results in italics indicate the best values; higher values are better