Skip to main content

Table 5 Results for the distribution learning Guacamol benchmarks

From: CReM: chemically reasonable mutations framework for structure generation

CaseMin increaseMax increaseMax replacementsValidityUniquenessNoveltyKL divergenceFrechet ChemNet Distance
CReM− 221001 ± 00.935 ± 0.0211 ± 00.443 ± 0.0230.021 ± 0.007
CReM− 22101 ± 00.942 ± 0.0081 ± 00.530 ± 0.0610.024 ± 0.034
CReM− 2251 ± 00.941 ± 0.0031 ± 00.572 ± 0.0380.044 ± 0.053
CReM− 2221 ± 00.950 ± 0.0021 ± 00.551 ± 0.0540.019 ± 0.018
CReM− 661001 ± 00.942 ± 0.0230.999 ± 00.541 ± 0.0560.018 ± 0.012
CReM− 66101 ± 00.924 ± 0.0101 ± 00.603 ± 0.0190.041 ± 0.045
CReM− 6651 ± 00.921 ± 0.0221 ± 00.584 ± 0.0340.038 ± 0.040
CReM− 6621 ± 00.935 ± 0.0091 ± 00.605 ± 0.0150.053 ± 0.050
CReM− 10101001 ± 00.918 ± 0.0191 ± 00.531 ± 0.0580.071 ± 0.027
CReM− 1010101 ± 00.907 ± 0.0220.999 ± 0.0010.622 ± 0.0110.030 ± 0.016
CReM− 101051 ± 00.875 ± 0.0251 ± 00.599 ± 0.0350.085 ± 0.056
CReM− 101021 ± 00.850 ± 0.0941 ± 00.590 ± 0.0640.006 ± 0.005
CReM− 1021001 ± 00.945 ± 0.0210.999 ± 00.550 ± 0.0370.016 ± 0.012
CReM− 102101 ± 00.950 ± 0.0081 ± 00.545 ± 0.0070.045 ± 0.010
CReM− 10251 ± 00.956 ± 0.0011 ± 00.533 ± 0.0730.048 ± 0.036
CReM− 10221 ± 00.962 ± 0.0061 ± 00.577 ± 0.0270.042 ± 0.037
SMILES LSTM*   0.95910.9120.9910.913
Graph MCTS*   110.9940.5220.015
AAE*   0.82210.9880.8860.529
ORGAN*   0.3790.8410.6870.2670
VAE*   0.8700.9990.9740.9820.963
  1. * Results were taken from the ref [38]