CReM: chemically reasonable mutations framework for structure generation

Table 5 Results for the distribution learning Guacamol benchmarks

Case	Min increase	Max increase	Max replacements	Validity	Uniqueness	Novelty	KL divergence	Frechet ChemNet Distance
CReM	− 2	2	100	1 ± 0	0.935 ± 0.021	1 ± 0	0.443 ± 0.023	0.021 ± 0.007
CReM	− 2	2	10	1 ± 0	0.942 ± 0.008	1 ± 0	0.530 ± 0.061	0.024 ± 0.034
CReM	− 2	2	5	1 ± 0	0.941 ± 0.003	1 ± 0	0.572 ± 0.038	0.044 ± 0.053
CReM	− 2	2	2	1 ± 0	0.950 ± 0.002	1 ± 0	0.551 ± 0.054	0.019 ± 0.018
CReM	− 6	6	100	1 ± 0	0.942 ± 0.023	0.999 ± 0	0.541 ± 0.056	0.018 ± 0.012
CReM	− 6	6	10	1 ± 0	0.924 ± 0.010	1 ± 0	0.603 ± 0.019	0.041 ± 0.045
CReM	− 6	6	5	1 ± 0	0.921 ± 0.022	1 ± 0	0.584 ± 0.034	0.038 ± 0.040
CReM	− 6	6	2	1 ± 0	0.935 ± 0.009	1 ± 0	0.605 ± 0.015	0.053 ± 0.050
CReM	− 10	10	100	1 ± 0	0.918 ± 0.019	1 ± 0	0.531 ± 0.058	0.071 ± 0.027
CReM	− 10	10	10	1 ± 0	0.907 ± 0.022	0.999 ± 0.001	0.622 ± 0.011	0.030 ± 0.016
CReM	− 10	10	5	1 ± 0	0.875 ± 0.025	1 ± 0	0.599 ± 0.035	0.085 ± 0.056
CReM	− 10	10	2	1 ± 0	0.850 ± 0.094	1 ± 0	0.590 ± 0.064	0.006 ± 0.005
CReM	− 10	2	100	1 ± 0	0.945 ± 0.021	0.999 ± 0	0.550 ± 0.037	0.016 ± 0.012
CReM	− 10	2	10	1 ± 0	0.950 ± 0.008	1 ± 0	0.545 ± 0.007	0.045 ± 0.010
CReM	− 10	2	5	1 ± 0	0.956 ± 0.001	1 ± 0	0.533 ± 0.073	0.048 ± 0.036
CReM	− 10	2	2	1 ± 0	0.962 ± 0.006	1 ± 0	0.577 ± 0.027	0.042 ± 0.037
SMILES LSTM^*				0.959	1	0.912	0.991	0.913
Graph MCTS^*				1	1	0.994	0.522	0.015
AAE^*				0.822	1	0.988	0.886	0.529
ORGAN^*				0.379	0.841	0.687	0.267	0
VAE^*				0.870	0.999	0.974	0.982	0.963

ISSN: 1758-2946