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

Table 2 Shingles statistics for the datasets at different stages of the workflow

Dataset	Size	Distinct shingles r1	Distinct shingles r2	Unique shingles r2	Distinct shingles r3	Unique shingles r3
QM9 step 3	122,227	229	28,053	7162	376,852	273,423
PC9 step 3	77,790	1295	39,725	18,718	223,127	158,226
OD9_0 step 3	184,158	1297	57,741	22,130	544,460	392,637
OD9_1 step 1	1,023,624	1007	642,265	282,311	4,568,964	3,675,203
OD9_1 step 2	854,059	3585	979,596	548,870	4,255,262	3,513,467
OD9_1 step 3	250,874	762	213,034	103,858	1,156,813	929,228
OD9 step 1	1,276,171	2447	691,715	301,669	5,156,545	4,064,788
OD9 step 2	1,088,773	3714	1,013,639	557,832	4,798,140	3,870,539
OD9 step 3	435,032	1563	250,163	116,483	1,665,725	1,293,995

ISSN: 1758-2946