Skip to main content

Table 5 Performance comparison of the Transformer models trained on different types of molecular pairs on the restricted intersection test set (numbers in bracket represent the absolute increase or decrease compared to the corresponding Transformer model performance on the original test set in Table 4)

From: Transformer-based molecular optimization beyond matched molecular pairs

Test set Type of molecular pairs where Transformer is trained Successful property constraints (%) Successful structure constraints (%) Successful property and structure constraints (%)
  MMP 65.71 (\(\uparrow\) 3.81) 91.68 (\(\uparrow\) 0.13) 61.82 (\(\uparrow\) 3.73)
  Similarity (\(\ge\)0.5) 55.55 (\(\uparrow\) 3.72) 84.47 (\(\uparrow\) 2.17) 48.97 (\(\uparrow\) 4.44)
Restricted Similarity ([0.5,0.7)) 50.17 (\(\uparrow\) 3.42) 68.66 (\(\uparrow\) 0.57) 35.28 (\(\uparrow\) 2.32)
intersection Similarity (\(\ge\)0.7) 65.39 (\(\uparrow\) 0.30) 81.49 (\(\downarrow\) 1.19) 55.55 (\(\downarrow\) 0.52)
  Scaffold 62.91 (\(\uparrow\) 1.38) 94.42 (\(\downarrow\) 0.90) 60.70 (\(\downarrow\) 1.01)
  Scaffold generic 59.07 (\(\uparrow\) 4.02) 96.14 (\(\uparrow\) 0.13) 57.68 (\(\uparrow\) 4.02)
  1. The extremes (best/worst performance or largest/smallest change) are highlighted in bold