The development of models to predict melting and pyrolysis point data associated with several hundred thousand compounds mined from PATENTS

Table 2 Analyzed sets of descriptors

Package name	Type of descriptors^a	Number of descriptors	Matrix size, billions	Number of descriptors after filtering	Non-zero values, millions	Sparseness^b
EFG	Binary	595	0.18	347	3.1	33
QNPR	Integer	1502	0.45	1040	6.3	49
MolPrint	Binary	688,634	205	197,367	8.1	7200
E-state count	Float	631	0.19	487	10	14
Inductive	Float	54	0.02	39	11	1
ECFP4	Binary	1024	0.31	1021	12	25
ISIDA	Integer	5886	1.75	2275	18	37
ChemAxon	Float	498	0.15	114	23	1.5
GSFrag	Integer	1138	0.34	469	24	5.7
CDK	Float	239	0.07	182	27	2
Adriana	Float	200	0.06	139	32	1.3
Mera, Mersy	Float	571	0.17	235	61	1.1
Dragon	Float	1647	0.49	911	183	1.5

^aThe dominating type of descriptors within the corresponding package
^bAverage number of zero entries per one non-zero value of the descriptor matrix

ISSN: 1758-2946