The influence of negative training set size on machine learning-based virtual screening

Table 2 The improvement of MCC values calculated for targets from the confirmatory tests

Target	ChEMBL class	Train/test	CDK FP/MCC
			SMO	NB	Ibk	J48	RF
D2	membrane receptor	310/1407	0.55	0.04	0.48	0.25	0.65
EGFR	enzyme/kinase	280/1303	0.35	0.09	0.50	0.41	0.61
Mu opioid	unclassified protein	270/1235	0.53	0.05	0.55	0.33	0.65
SERT	transporter	390/1822	0.25	0.03	0.62	0.40	0.55
Estrogen α	transcription factor	133/614	0.47	0.10	0.46	0.28	0.67
AChE	enzyme/hydrolase	162/743	0.58	0.10	0.41	0.23	0.69
Factor Xa	enzyme/protease	530/2439	0.54	0.02	0.58	0.39	0.67
Thrombin	enzyme/protease	370/1691	0.59	0.04	0.58	0.27	0.66
PDE5	enzyme/phosphodiesterase	152/695	0.56	0.01	0.36	0.28	0.60
Renin	enzyme/protease	340/1556	0.46	0.06	0.59	0.33	0.65
Glucocorticoid	transcription factor	236/1084	0.62	0.03	0.56	0.31	0.74
CRF1	membrane receptor	200/914	0.59	0.03	0.46	0.34	0.74

The table shows the changes in MCC for a particular ML method obtained between experiments with the lowest and the highest ratio of negative to positive training examples.

ISSN: 1758-2946