Skip to main content

Table 3 Scaffolds, IFG and shingles statistics (averaged over 10 runs) for the QED goal-directed experiment for different descriptors and different weights for the entropy term

From: Scalable estimator of the diversity for de novo molecular generation resulting in a more robust QM dataset (OD9) and a more efficient molecular optimization

Optimized descriptor Entropy weight Mean QED Distinct scaffolds Distinct IFG Distinct checkmol Distinct shingles r1 Distinct shingles r2 Distinct shingles r3
None (i.e. QED only) 0 0.944 196 259 25 156 1854 5579
IFG 0.1 0.948 329 467 27 230 2570 6719
  1 0.947 670 859 44 375 4741 9627
  10 0.917 771 1221 63 684 6901 12,149
  100 0.048 648 2526 79 1302 12,902 20,799
  1000 0.034 607 2479 76 1314 12,714 20,586
Checkmol 0.1 0.948 265 375 32 191 2197 6141
  1 0.947 372 423 59 284 2985 7113
  10 0.925 415 470 106 365 3382 7596
  100 0.391 561 799 137 493 4950 11,006
  1000 0.074 545 929 140 604 5735 12,379
Shingles r1 0.1 0.948 466 600 38 451 3674 7765
  1 0.945 718 919 53 801 5978 10,423
  10 0.767 745 1176 76 2306 10,485 15,164
  100 0.036 798 1038 58 4681 23,328 30,813
  1000 0.036 802 1043 60 4802 23,305 30,803